-- GiuliaTucci - 06 Aug 2024

Uncovering visual narratives about human mobility on Telegram

Participants*: Aliki Livada, Benedetta Riccio, Daniel Lundgaard, Heidi Ippolito, Riccardo Freschi, Sara de Andrade Silva, Tahereh Aboofazeli

*in alphabetical order

Designer: Benedetta Riccio

Data Engineer: Giosuè Ruscica

Facilitators: Bia Carneiro, Giulia Tucci, Giosuè Ruscica

Case studies: Conspiracist perspective: Heidi Ippolito and Aliki Livada | Nationalist perspective: Sara de Andrade Silva | Climate action perspective: Daniel Lundgaard | African continent: Giulia Tucci and Bia Carneiro | Afghanistan and Iran: Tahereh Aboofazeli | Ukraine and Visegrad: Riccardo Freschi

Links

• Dataset: https://doi.org/10.5281/zenodo.12663541

• Poster: http://dx.doi.org/10.13140/RG.2.2.25719.02721

1. Introduction

Human mobility is the term that broadly encompasses different types of movement by people. It includes international migration, as well as internal and cross-border movements, voluntary and forced migration, refugee flows, internal displacement, planned relocations and resettlements. Voluntary and forced immobility are also considered part of the phenomenon.

According to the International Organization for Migration (IOM), the current global estimate is that there were around 281 million international migrants in the world in 2020, the equivalent of 3.6 per cent of the global population. (IOM, 2022). The UN Agency for Refugees reported that, at the end of 2023, 117.3 million people worldwide were forcibly displaced as a result of persecution, conflict, violence, human rights violations and events seriously disturbing public order (UNHCR, 2024) . These figures and recent upward trends evidence the relevance of human mobility to social dynamics

However, human mobility is not uniform across the world, but is multicausal and shaped by social, economic, geographic, demographic and other factors that result in distinct mobility (and immobility) patterns (Savelli et al., 2023). Addressing these dimensions requires comprehending the various factors contributing to them, from diverse perspectives.

Given the growing significance of digital/platform-based information flows, a major advantage of using digital data to analyse human mobility-related aspects is the potential to generate additional streams of information and insight, which complement traditional data sources and research methods (Carneiro & Tucci, 2024). While governments and international agencies have progressed in collecting human mobility statistics, traditional sources have several limitations due to their cost, coverage and timeliness. In addition, they are often unable to capture the underlying nuances driving migration and displacement (both international and internal), such as factors influencing migrant decision-making, and public opinion or discourses on the issue.

Scholarly literature is cognizant of the interconnections between digital technologies and the movement of people, with several studies on the use of information and communication technologies within human mobility contexts, as well as in attempting to quantify and qualify human mobility patterns (Carneiro & Tucci, 2024).

Social media data has been widely used to gain insights into social phenomena. In the digital environment, Telegram excels with its expansive, encrypted conversation capabilities, making it an invaluable tool for studying communication dynamics. Initially a messaging app, Telegram has integrated features that transformed it into a hybrid platform (Tucci, 2023). It enables information dissemination on a massive scale through broadcasting channels and supergroups, catering to large audiences. It also has become an arena for the deplatformed (Rogers, 2020)— those ejected from mainstream social media for policy violations.

These aspects enhance Telegram's research potential, providing insights into the raw exchange of ideas, including those marred by disinformation and hate speech (Mitts, 2022; Rogers, 2020). This unique blend of freedom and risk underscores the complexity of balancing open discourse with digital responsibility, and positions Telegram at the forefront of discussions on the impact of deplatforming in the digital sphere. This makes Telegram well-suited for observing the organic development and dissemination of narratives related to human mobility.

This research project aimed to investigate human mobility visual narratives that emerge from Telegram images' visual vernaculars (Colombo et al., 2023; Gibbs et al., 2015). After creating a curated list of Telegram groups and channels, TeleCatch (Ruscica et al., 2023) was employed to construct datasets for thematic case studies. TeleCatch is an open-source tool developed by the Digital Innovations team at CGIAR FOCUS Climate Security to visualise, filter and extract data from a list of Telegram groups and/or channels. The application of Artificial Intelligence tools to analyse image collections generated from these datasets uncovered a range of visual narratives depicting human mobility across the case studies.

Context

The Consultative Group for International Agricultural Research (CGIAR) is a global research-for-development partnership that aims to deliver science and innovations for sustainable and resilient food, land, and water systems. Its Research Initiative on Fragility, Conflict, and Migration (FCM) aims to enhance this resilience in fragile and conflict-affected settings, where human mobility-related challenges are prevalent.

Displacement and forced migration often result from and further escalate the humanitarian and development challenges fragile and conflict-affected settings (FCASs) face. 80% of forcibly displaced individuals suffer from acute food insecurity and high levels of malnutrition. 83% of the world’s international refugees, many of whom have been displaced for years, even decades, are hosted by low and middle-income countries (Kosec et al., 2023).

The FCM team works closely with governments and other partners on local, national, and regional levels to generate evidence on effective policies and programming before, during, and after shocks and crises that have implications to human mobility. It conducts research on responses to conflict and displacement, contributing to improving understanding of conflict hotspots and migration corridors; improving understanding of the dynamics of human mobility and conflict for more effective, context-specific programming; and leveraging sustainability principles to ensure better alignment of humanitarian, development and peace interventions. Within FCM, a digital innovations thematic area has been applying digital methods to digital platform narratives about human mobility, to uncover migrant communities’ concerns and drivers to decision-making, especially focusing on displacement and interlinkages to climate, peace and security.

Online issue mapping was applied to Twitter data to assess policymaker discourses around the climate security-migration nexus in Nigeria (Carneiro et al., 2022), as well as to capture diverging priorities around this nexus by national and international actors in Mozambique and Jordan (Carneiro et al., 2023b, 2023a). A working paper analysing public conversations about human mobility in African Telegram groups and channels (Carneiro & Tucci, 2024) was the starting point for this project.

2. Research Questions

The project aims to investigate three research questions:

1. How is human mobility represented on Telegram message content?

2. How do these representations change across thematic groups and geographies?

3. How does AI interpret human mobility visual vernaculars?

3. Methodology

a) Dataset Creation

To create working datasets for analysis of human mobility discourses on Telegram groups and channels, the TeleCatch tool was employed. The dataset creation steps are illustrated below. Based on the selection of "Human Mobility" as the topic for research, a taxonomy was devised to identify relevant keywords for data collection, ensuring that the terms were comprehensive and targeted for the study. Concurrently, lists of relevant Telegram groups and channels were curated based on the themes of the case studies defined by the project participants. These curated lists were loaded into TeleCatch and their content was filtered through queries using the taxonomy. Data were then aggregated and cleaned to create six working datasets.

The thematic framework of analysis for the case studies is represented below. This framework includes diverse geographies and perspectives to analyse human mobility discourses. Considering the effects of right wing extremism in migration perceptions in the global north, two such perspectives were analysed - conspiracist and nationalist propagandist - as well as a counterpoint from climate activist communities. The analyses also zoomed into three geographies with distinct experiences in human mobility: the African continent, with its complex displacement and migration dynamics; movement of Ukrainian refugees into the countries of the Visegrad group (Czech Republic, Hungary, Poland and Slovakia); and the movement of Afghanistani people into Iran.

The keywords related to human mobility employed to query the data via TeleCatch:

b) Methodological strategy

To explore the research questions , for each case study we followed the two paths:

1. Cloud Vision web entities analysis: First, we used Memespector GUI (Chao, 2023) to access Google Computer Vision API (Google Cloud, n.d.) and obtain the web entities annotation for all images related to human mobility extracted from Telegram. Then, we used Gephi (Bastian et al., 2009) to generate a network of web entity co-occurrences.

2. GPT analysis of image walls colour clusters: We loaded the image folder containing the images extracted from Telegram messages on the ImageSorter software (Visual Computing Group 2018) to obtain an image wall for each case study. We sorted the images by colour, qualitatively analysed the colour clusters, and produced screenshots of those identified as most relevant. Then, we copied and pasted the screenshots of the entire image walls and of the most relevant colour clusters on Open AI's generative model, Chat GPT (OpenAI, 2023), using the following prompt:

4. Results and discussion

a) Conspiracist perspective

Researchers’ motivation: Heidi Ippolito was motivated to join this case study because it overlaps with her PhD research on QAnon and conspiratorial communities. Up to this point, Heidi’s research has been primarily observational, so she was keen to apply the Telecatch tool (as well as tools like Gephi, Memespector, ​​and PicArrange) to create and analyse datasets. Aliki Livada's motivation for participating in the conspiracy subgroup stems from her interest in American politics, specifically the views of Donald Trump supporters on immigration, particularly in light of the upcoming American elections later this year.

Telegram groups and channels list curation and TeleCatch keyword querying

• For the Telegram list creation, we started with a preliminary list of QAnon channels previously collected by Heidi. We then expanded this list by using search terms (e.g., WW1WGA, QAnon, The Storm, etc.) in TGStat and snowballing with “Similar Channels” suggested by Telegram in relevant channels.

• Our final list of Telegram channels included 65 primarily English-language channels (unsurprising, due to the US-origins of QAnon), but the list also includes 9 German-language groups, as well as 2 Hebrew, 2 Russian, 1 Danish, 1 French, 1 Italian, 1 Spanish, 1 Ukrainian, and 1Uzbek.For the keyword querying in TeleCatch, we queried all English terms from the taxonomy and included additional human mobility keywords that frequently appear in these types of online conspiratorial communities (e.g., border, “build the wall”, “human trafficking”, “illegal aliens”, illegals, invaders). Considering the non-English language groups, we queried the keywords in German that yielded the largest results in English, including grenze [border], menschenhandel [human trafficking], illegale [illegals]. A total of 29 keywords were used.

• The final image dataset comprises 9,540 images based on the total message data collected through TeleCatch, which we then used to generate the visual analysis and network analysis.

Main Findings

General narratives from Chat GPT

Overall themes: fear, surveillance, migration, disinformation

• Lots of text images, often quotes, explainers, and proofs that cohere conspiratorial narratives (“The presence of text-heavy images with green backgrounds or highlights may point to informational content, warnings, or instructions related to the conspiracy theories.”)

• The images convey a view of human mobility as scary, urgent, political, and heavily surveilled.

• The green cluster is comprised mostly of maps, military personnel, and high angle (drone) shots of large groups of people. Rather than a celebration of nature, these images of the “outdoors” convey alarm and discomfort: “images of people moving en masse, border scenes, and cultural integration or clash” as well as “Nature and Outdoors, Military and Law Enforcement, Protests and Gatherings, Political Figures”

• The red and yellow clusters contain high emotional content that includes urgency, breaking news, and political figures.

Methodological Observations

• QAnon and conspiracy-related Telegram channels were easily accessible, and it might be interesting to approach the same research questions using non-QAnon conspiracy perspectives (e.g., groups that are less US-centric, New Age “conspirituality” groups, etc.)

• We made adjustments to keywords in order to retrieve a larger set of results (e.g., “sex traffic” did not yield high results, so we used “sex trafficking”)

Image Clusters and Chat GPT Analysis

• Chat GPT gave decent overviews of the colour clusters, but some of the analysis seemed vague or overreaching. For example, the AI’s interpretation “The varied colour patterns might also indicate the spread of disinformation. Different colour clusters can represent different types of content such as fake news, conspiracy theories, and propaganda” seems like a leap; different colours can mean all sorts of things, and not necessarily disinformation. In this case, it is important to comb through the results for misleading analysis from AI tools.

Google Cloud Vision Web Entities and Network Analysis Visualization:

• We used the web entities determined by Google Cloud Vision from the corpus of images to produce an entity co-occurrence network and visualise connections between image features.

• This visualisation shows strong US-centric themes, with “United States”, “Politics”, “Immigration”, “Border,” and “US President” as some of the largest overall connections. This is unsurprising, as most of the conspiracy channels from Telegram are English language and (likely) based in the US.

b) Nationalist perspective

Researchers’ Motivation: In the past few years, nationalist parties seem to be making a comeback. From the Eurosceptic populist AfD in Germany to the anti-immigration Chega movement in Portugal, the far-right has been increasingly relying upon nationalist inspiration to appeal to the masses. In this context, migration has been posed as a threat, and used by these movements as a scapegoat for explaining a set of issues, from violence to unemployment. This exploration sought to uncover how migration was represented in nationalist Telegram groups in a comparative perspective using three different countries: the United States, Italy, and Portugal.

Telegram groups and channels list curation and TeleCatch keyword querying

From “official” politics/parties channels on Telegram - the official Donald Trump account for the US, Giorgia Meloni for Italy, and the CHEGA channel for Portugal - we used the “Similar Groups” tab to find related groups to curate our list. The saturation point was reached when I started to be redirected to the same groups and/or really niche or small groups. For Italy and Portugal, this happened at around 7 groups, but I did not reach the saturation for the US context. Seeking to maintain an equal sample size for each country, I arbitrarily stopped collecting groups in the American scenario.

Using the pre-determined taxonomy for human mobility, we used Telecatch to search within the groups and collect all relevant messages to create a dataset containing images and image texts.

Using the images, we created an image wall to cluster images together by color (ImageSorter) and similarity (PixPlot), from which we gathered the first insights using generative AI (ChatGPT).

In the next step, we used Google Vision AI to interpret and describe the images found in the Telegram groups. The descriptions were then used to create a network graphic to understand how topics discussed were related across groups.

Main Findings

Nationalism and misinformation/disinformation on human mobility

1. Number of Telegram Groups: 25

2. Telegram Group Search: From “official” politics channels to related ones using the “similar groups” tab.

3. Additional TeleCatch Query Keywords: the same list was translated into Italian and Portuguese and used to capture messages in all groups.

4. Number of Images: 7,184 (analysis was done on 2043 images)

5. Languages: English, Italian and Portuguese

General narratives from Chat GPT

Emergent Narratives

1. Activity and Crowd Dynamics:

   • The image wall shows a significant number of images depicting large crowds, gatherings, and events. This suggests a focus on public activities, social movements, or mass events, which are central to the theme of human mobility.

2. Geographical Diversity:

   • There are varied landscapes and settings indicating different geographical locations. This points to the diverse contexts in which human mobility occurs, spanning urban, rural, and natural environments.

3. Social Interactions:

   • Numerous images show people interacting in various contexts such as protests, celebrations, or everyday activities. This highlights the social aspect of human mobility, emphasizing how people come together for different purposes.

4. Infrastructure and Transportation:

   • Several images include transportation modes like cars, bikes, and public transit. This indicates the importance of infrastructure in facilitating human movement and the different means by which people travel.

5. Documentation and Media Coverage:

   • There are images of documents, media reports, and social media posts. This suggests the role of media in documenting and influencing perceptions of human mobility, as well as the dissemination of information related to it.

Humanitarian/migration agencies clusters

1. Humanitarian and Refugee Context:

   • Multiple images prominently feature the UNHCR (United Nations High Commissioner for Refugees) logo and related activities. This suggests a strong focus on humanitarian issues, refugee movements, and international support for displaced populations.

2. Political Figures and Speeches:

   • There are several images of political figures giving speeches, which indicates discussions or announcements about policies, humanitarian aid, or international relations. This includes figures like Joe Biden and other prominent leaders.

3. Sea Travel and Migration:

   • Many images show people traveling by sea, lifeboats, and crowded boats. This highlights the perilous journeys undertaken by migrants and refugees, particularly those crossing seas to reach safer destinations.

4. Rescue and Aid Operations:

   • Visuals of rescue operations and aid distribution are evident, indicating the efforts made by international organizations and governments to assist those in need. Images of coast guards, aid workers, and medical assistance are present.

5. Flags and National Symbols:

   • Various national flags and symbols point to the involvement of different countries in addressing human mobility issues, suggesting a global perspective and international cooperation.

6. Maps and Geographic Focus:

   • The presence of maps indicates geographical tracking of migration routes or areas affected by mobility. This provides context to the regions from which and to which people are moving.

7. Media and Communication:

   • Screenshots from news broadcasts, social media, and other communication platforms show the role of media in documenting and spreading information about human mobility and related crises.

8. Healthcare and Medical Images:

   • Images related to healthcare, such as X-rays and medical workers, imply concerns about the health and well-being of migrants and refugees, including the medical challenges they face during and after their journeys.

The light blue cluster revolves around themes of humanitarian aid, political discourse, and the challenges faced by refugees and migrants. It highlights the global effort to manage and support displaced populations, with a strong emphasis on sea migration and rescue operations. The involvement of international organizations like the UNHCR and various countries' governments is evident, showcasing a multi-faceted approach to addressing human mobility crises.

Political figures, leaders and media coverage

1. Political Figures and Leadership:

   • A significant number of images feature political leaders and figures, including Joe Biden, Donald Trump, and other prominent politicians. This indicates a focus on political discourse, leadership, and decision-making processes related to human mobility and other related issues.

2. Legislative and Judicial Settings:

   • Many images depict scenes from legislative bodies, courtrooms, and official hearings. This suggests themes of policy-making, legal proceedings, and governance, highlighting how laws and policies impact human mobility.

3. Protests and Public Gatherings:

   • There are images of protests, public gatherings, and social movements. These visuals point to civil society's role in advocating for or against policies affecting human mobility, emphasizing the social and political activism surrounding these issues.

4. Media Coverage and Interviews:

   • Numerous images show media personalities, interviews, and news coverage. This underscores the role of media in shaping public perception and informing the public about political and social issues related to human mobility.

5. Transportation and Migration Events:

   • Some images depict transportation hubs, like airports, and scenes of migration events. This highlights the actual movement of people, the logistics involved, and the often complex circumstances of such mobility.

6. Law Enforcement and Security:

   • There are visuals of law enforcement, security measures, and police presence, indicating the security aspects related to human mobility. This could encompass border security, crime prevention, and maintaining order during mass movements.

7. Community and Personal Interactions:

   • Several images show personal interactions, community engagement, and everyday life, indicating the impact of political decisions on individuals and communities. This personalizes the broader political and legal narratives.

The brownish cluster centers around the political, legal, and social dimensions of human mobility. It highlights the roles of political leaders, legislative and judicial actions, and media coverage in shaping policies and public opinion. Additionally, it touches on the community and individual impacts, protests, and the security measures associated with human movement. This cluster provides a comprehensive view of how governance, activism, and media interplay in the context of human mobility.

Humanitarian agencies/migration images

The blue cluster of images appears to focus heavily on themes related to immigration enforcement, refugee situations, and the response of various authorities and organizations. Here is a detailed breakdown of the narratives and content within this cluster:

1. Government and Political Figures:

   • Several images feature prominent political figures (such as Joe Biden and Justin Trudeau), indicating discussions or actions taken by governments regarding immigration policies and refugee crises. This suggests a narrative of political leadership and decision-making in response to human mobility issues.

2. Law Enforcement and Border Control:

   • Numerous images depict law enforcement agencies, such as border patrol, police, and immigration enforcement officials. This highlights the emphasis on border security and the enforcement of immigration laws, which is a key aspect of nationalist approaches to human mobility.

3. Refugees and Migrant Camps:

   • Images of refugee camps, crowded boats, and groups of migrants emphasize the humanitarian aspect of human mobility. This showcases the challenging conditions faced by migrants and refugees, and the large-scale movements of people fleeing crises.

4. International Organizations:

   • The presence of UNHCR (United Nations High Commissioner for Refugees) logos and personnel underscores the role of international organizations in addressing refugee and migration issues. This highlights the global nature of the refugee crisis and the efforts of international bodies to provide aid and support.

5. Rescue and Humanitarian Efforts:

   • Several images show rescue operations at sea, crowded boats, and refugees being helped by humanitarian workers. This illustrates the life-saving efforts undertaken by various organizations to assist migrants in perilous situations.

6. Media and Reporting:

   • Images from news reports and media coverage indicate the significant attention these issues receive from the media. This reflects how media portrayal influences public perception and policy decisions regarding immigration and refugee matters.

7. Detention and Deportation:

   • Images of detention centers, deportations, and law enforcement interactions with migrants suggest the more punitive aspects of nationalist policies on human mobility. This includes the processes of detaining and deporting undocumented migrants.

8. Statistics and Data:

   • Some images include charts or statistical data, likely reflecting the scale of migration or the impact of policies. This suggests a narrative grounded in empirical evidence and data-driven discussions on the topic.

This cluster predominantly revolves around the following themes:

• Government Response: Actions and statements by political leaders regarding immigration policies.

• Enforcement: Border security, law enforcement, and the role of police in managing immigration.

• Humanitarian Crisis: The plight of refugees and migrants, showcasing the humanitarian response required.

• International Involvement: The involvement of international organizations like the UNHCR in addressing these issues.

• Media Influence: The role of media in shaping the narrative and informing the public about these issues.

• Statistical Analysis: Use of data to highlight the scale and impact of human mobility.

Overall, the cluster presents a multifaceted view of human mobility within the nationalist context, balancing the themes of security and enforcement with humanitarian concerns and international cooperation.

Mass migration narratives

The grey cluster of images appears to focus heavily on themes related to large-scale migration, border control, detention facilities, and the human impact of migration policies. Here’s a detailed breakdown of the narratives and content within this cluster:

1. Mass Migration:

   • Many images depict large groups of people moving together, often on foot, through rural and urban landscapes. This emphasizes the sheer scale of migration events and the collective movement of people, often fleeing from crises or seeking better opportunities.

2. Detention Facilities:

   • Several images show the interior and exterior of detention centers, with people held in confined, often overcrowded conditions. This highlights the response to mass migration from a detention and control perspective, underscoring the challenges of managing large influxes of migrants.

3. Protests and Unrest:

   • Images of protests, clashes with law enforcement, and civil unrest suggest a narrative of resistance and conflict related to immigration policies. These visuals reflect the tensions and opposition to both the treatment of migrants and the policies being enacted.

4. Border Security and Enforcement:

   • The presence of law enforcement, military personnel, and border fences or walls points to the focus on securing borders and controlling the movement of people. This aligns with nationalist priorities of maintaining strict border controls.

5. Humanitarian Conditions:

   • Images of makeshift camps, people living in dire conditions, and humanitarian aid efforts highlight the human cost of migration and the need for support. This provides a counterpoint to the enforcement narrative, showcasing the plight of migrants.

6. Political Figures and Policies:

   • Like the blue cluster, this set also features political leaders (such as Joe Biden) and media coverage of policy debates. This indicates the ongoing political discourse around immigration, highlighting key figures and their stances.

7. Transportation and Routes:

   • Images of migrants using various forms of transportation, including boats, trucks, and on foot, depict the perilous journeys undertaken by migrants. This illustrates the lengths to which people go to reach their destinations, often facing significant dangers.

8. Media and Public Perception:

   • The inclusion of news reports, interviews, and public statements highlights the role of media in shaping the narrative around migration. This reflects how public perception is influenced by the way these events are reported.

The grey cluster predominantly revolves around the following themes:

• Mass Migration: Large groups of migrants moving together, highlighting the scale and urgency of migration events.

• Detention and Control: The use of detention facilities and the enforcement of immigration laws, focusing on the control aspect.

• Protests and Resistance: Civil unrest and opposition to immigration policies, showcasing the conflict and resistance to current measures.

• Humanitarian Issues: The challenging conditions faced by migrants and the humanitarian efforts to support them.

• Political and Media Influence: The role of political leaders, policies, and media coverage in shaping the migration narrative.

Overall, the grey cluster presents a comprehensive view of human mobility within the nationalist context, balancing enforcement and control with the humanitarian aspects and the significant political and media attention these issues receive.

General narratives from Google Cloud Vision

First of all, we have to take into account the approach used to create the graphic: after collecting the data from the telegram groups, we extracted the images and uploaded them to Google Vision's AI to describe them. We are therefore working with broad categories of images circulating in these groups with immigration-related questions. Therefore, some of the results are expected, such as the fact that "immigration" and "human migration" are the largest nodes in the graph. At the same time, expected but not necessarily useful descriptions emerge: words like "screenshot", the Internet image bank "Getty Images" and "photo caption" are relatively frequent in the dataset.

Leaving these aside, some words can be used as indexes for discussions between these groups, and here we must consider academic knowledge on the subject. When we look qualitatively at the images, we notice that words like "rape" and "death" are frequently present in the news, and denounce crimes committed by migrants, in a narrative that tries to link migrants to violent crimes. This finding is consistent with the literature (Kaufmann, 2017; Yilmaz, 2012) on far-right narratives about migrants, which often resort to widespread moral terrorism among these groups.

Although we haven’t had the chance to analyze closely the link between the images and texts circulating in these groups, some images such as official humanitarian agencies' logos and spokespeople, we can hypothesize that these groups rely heavily upon mainstream media images to craft their own texts. At the same time, these images, taken from recognized media portals might help them to create a truth varnish, which might improve their appearance as a credible source among the channel’s subscribers.

Images containing large numbers of people on boats or roads might help in constructing the “white replacement” narrative, a widespread conspiracy theory among right-wing groups that uses pseudoscientific sources and visual appeals to portray immigration as a threat (Ekman, 2022). Finally, it is interesting to note how migration is a relevant topic among nationalist Telegram groups. This suggests these groups engage in a “we” vs “they” kind of narrative, observed in the classical literature on populism (Mouffe, 2011). Theory-wise some possibilities emerge, the Strong Program in Cultural Sociology (Alexander and Smith, 2010) being one of the promising frameworks. Relying on a strong hermeneutical approach, the Strong Program can allow us to analyze the narrative construction in these groups, understanding how words, images, and meanings are correlated to create a coherent narrative that is highly appealing for some segments in different national contexts. At the same time, the data gathered opens the possibility to compare and understand how populism works, providing empirical data to test the broad theoretical literature on the topic. However, because of the exploratory nature of this investigation, these analyses remain beyond the scope of this report.

c) Climate action perspective

Researchers’ motivation: Climate change is a global issue, with rising temperatures and rapidly changing weather conditions affecting everyone around the globe. Climate-induced environmental changes, such as rising sea levels, extreme weather events, and droughts, increasingly force people to migrate from affected areas, often leading to significant socio-economic and geopolitical challenges (McLeman, 2019). As such, to uncover the visual narratives about human mobility on Telegram, it is important to also consider how visual narratives about human mobility emerge in relation to climate action.

Telegram groups and channels list curation and TeleCatch keyword querying

• For the Telegram channel list creation, I first started out by searching for climate action and climate activism related words on the Telegram web app and reviewing the channels and groups that appeared. This initial list of keywords included: “climate action”, “climate activism”, “environmental action”, “environmental activism”. Expanding on this list, I reviewed channels and groups identified using broader search terms, including “Climate”, “activism”, “action”, “sustainability” and “environment”.

• To further expand the dataset, I also reviewed channels and groups related, referring to or moderated by known climate action/activism organisations, including “Extinction Rebellion”/”XR”, “Greenpeace”/”GP”, “Fridays For Future”/”FridaysForFuture”/”FFF”, “Cool earth”, “World Wildlife Foundation”/”WWF”.

• Throughout this process, all channels and groups were reviewed and shared content was qualitatively analysed for two reasons: (1) to ensure the content of the group/channel reflected a desire to stop climate change, not discuss or spread climate scepticism, and (2) to identify potentially relevant new groups and channels mentioned in discussions within these groups/channels.

• The review of ‘other’ channels resulted in the identification of new relevant channels such as “PlanetFreaks”, “PollutionFacts”, “Ecocide”, “MissionPlaneta”.

• For all search words, the search included both the word-query, and the hashtagged word query, e.g., “climate action” and “#climateaction”, however, for the majority of the searches the results were very similar.

• In total, this process produced a list of 20 Telegram groups and channels

• TeleCatch was used to query the groups and channels for the human mobility taxonomy. The posts within the channels that contained any keyword from the taxonomy were then downloaded, including media content embedded in the posts. The final image dataset comprises 39 images.

Main findings

First and foremost, the most interesting (non) finding is that climate action rarely appears in relation to human mobility (on Telegram). Despite ongoing discussions of climate-induced migration as part of the broader climate action-conversation, human mobility (as defined by the keyword queries) does not appear to be a central topic in relation to climate action on Telegram. This is both based on the low number of related channels identified, but also that TeleCatch returned relatively few results when querying for the keywords, resulting in a comparatively low sample of 39 images.

With that said, the AI assisted analysis (ChatGPT) of the images does reveal a link between human mobility and climate action, including the impact of climate change on human mobility, e.g. how flooding and droughts leads to increased human mobility, this is in particular seen in the images illustrating scorched lands or flooded areas illustrating the severe effects of climate change on habitability and subsequently contributing to human mobility. However, the analysis of the images also reveal a broader focus, with the images found by the queries highlighting the environmental impact of climate change beyond human mobility, including impact on wildlife that are “migrating” as well.

Additionally, the narratives conveyed in the images relate to broader questions about infrastructure and development, and a more scientific approach with graphs and analytical insights as often seen in debates about climate change.

Finally, the analysis of the images also reveals a strong emphasis on community and solidarity, linking back to broader tendencies within the climate action conversation about this being a global challenge, necessitating a collective effort, both to combat climate change more broadly, but also specifically the implications in human mobility as a result of climate change, and how this also requires collective effort.

Reviewing the network based on the descriptions generated by the Google Cloud Vision, I find a strong emphasis on “Immigration”, which reflects the overall focus of this project. Naturally, due to the theme of the category, “climate change” also appears as a central theme in the images. Finally, the network analysis highlights a “descriptive”/“genre”-cluster, which highlights technical aspects of the images, e.g., that an image includes text, uses photo caption, resembles a document, contains a screenshot etc.

Discussion

How is human mobility represented on Telegram message content?

Within the comments and images related to human mobility shared on Telegram, human mobility is represented as a consequence of climate change. While climate migration is recognized as an important issue, human mobility is more commonly presented as a consequence of climate change, and used to emphasise the importance of fighting climate change, while climate migration is only to a lesser extent recognized as a key issue in itself. Additionally, it is worth noting that “migration” within the climate action debate also includes the migration of animals and wildlife, as their usual habitats are affected by climate change, but, again, the key focus is on climate change causing this migration, and thus climate change is a problem, not that the migration itself is a challenge.

However, because of the relatively low number of relevant channels identified, and subsequently the low number of both comments and images returned by the queries, it is relevant to question whether this is because Telegram is used less frequently than other platforms to discuss climate action, or because human mobility is not a central topic within the climate action debate. As the findings suggest, human mobility is a relevant topic within the climate action debate on Telegram, but climate action-debates more broadly only appear infrequently on the platform. This can be explained by how Telegram often is used for counter-discourses, and thus, the debate on climate change around which there is a widespread consensus, might be more commonly found on other platforms. This is echoed by the fact that well-known climate action organisations such as Extinction Rebellion only lists other social media platforms, including Instagram, Facebook, X, TikTok and YouTube as their primary platforms, but not Telegram, illustrating that this is not a platform they are focused on in their climate action efforts.

How does AI interpret human mobility visual vernaculars?

Overall, the AI performs well with regards to interpreting human mobility visual vernaculars in the context of climate action discussions. Naturally, there are limitations with using different AI models to analyse the “narratives” emerging from the way that images are clustered (by another program), but the qualitative review of the images highlights that the AI model identified commonalities and narratives across the images that match with the qualitative review. This highlights the potential of AI for uncovering visual narratives. Naturally, one should always be careful about relying completely on AI models, but they do provide opportunities for larger scale visual narrative analysis that should not be overlooked.

Similarly, for the results from Google Cloud Vision, the AI model identified patterns in images relevant for further analysis. Nothing new or groundbreaking emerged from this part of the analysis, as it was to be expected that immigration and climate change would appear as central themes. However, using network analysis to reveal patterns in the web entities of the images presents interesting opportunities for future research as it is a novel way of visualising (and analysing) patterns in content of images, as well as the centrality of certain themes.

d) African continent

Researchers’ motivation: Studying human mobility in the African continent is crucial due to its complex and dynamic nature, which is influenced by a combination of social, economic, political, and environmental factors. Africa is home to a diverse range of migration patterns, including internal displacement, cross-border movements, and international migration. Understanding these patterns is essential for addressing challenges such as economic inequality, political instability, and climate change impacts, which are often drivers of migration. Additionally, studying African migration provides insights into the cultural and historical contexts that shape the movement of people, offering valuable perspectives for policymakers and scholars aiming to develop more effective and humane migration policies globally. This research can contribute to a deeper comprehension of global migration trends, fostering international cooperation and support for sustainable development in the region.

Telegram groups and channels list curation and TeleCatch keyword querying

Groups and channels in English identified as sharing information relate to the African continent context. A total of 28 groups and channels were found using Telegram's search function, and to ensure they represented the wide continent, their descriptions and content were qualitatively assessed. Working with this collection in TeleCatch, the migration taxonomy terms in English were applied to filter for messages of interest.

• Number of Telegram Groups and channels: 28

• Number of Images: 3073

• Languages: English

Main findings

Chat GPT interpretations

African continent image wall: observations from entire wall

1. Colour Distribution:

   • The image wall is densely packed with images, with clear variations in colour density across different sections.

   • The left side of the wall has a more earthy tone with browns, greens, and yellows, suggesting natural landscapes, rural settings, or agricultural activities.

   • The right side transitions into a mix of grayscale and more vibrant colours, indicating urban settings, night scenes, or areas with significant human activity.

   • There are patches of brighter colours scattered throughout, which could represent cultural events, festivals, or other vibrant gatherings.

2. Image Clustering:

   • The clustering of images seems to follow a gradient, with more uniform and similar images grouped together.

   • The central portion shows a mix of colours and images, possibly indicating a transition zone or areas where different activities or themes intersect.

   • The top and bottom sections seem to be more uniform in their respective colour schemes and themes.

3. Themes and Narratives:

   • Left Side (Natural and Rural): Images here likely depict rural life, natural landscapes, agriculture, and possibly wildlife. This could narrate the daily life, struggles, and beauty of rural communities in Africa.

   • Right Side (Urban and Vibrant): This area seems to capture urban life, cultural events, night scenes, and other human-centric activities. It might tell stories of city life, modernization, challenges of urbanisation, and the vibrancy of African cultures.

   • Central Area (Intersection): The mix here suggests zones of transition, possibly areas undergoing change or regions where rural and urban elements coexist. This could reflect migration, development projects, or social and economic changes.

Observations from colour clusters

Blue Cluster

1. General Themes:

   • Water and Marine Activity: Many images feature boats, the ocean, and people engaged in activities on or near water. This suggests themes of travel, migration, fishing, and possibly maritime security or rescue operations.

   • Conferences and Political Figures: There are numerous images of people speaking at conferences or political events, indicating discussions around policy, international relations, or economic issues.

   • Flags and National Symbols: Various national flags and symbols are visible, hinting at themes of national identity, sovereignty, and international cooperation.

   • Economic and Financial Institutions: Several images reference banks or financial institutions, pointing towards discussions on finance, trade, and economic development in Africa.

2. Specific Observations:

   • Migration and Refugees: The presence of boats and crowded vessels likely depict scenes of migration, refugees, and possibly rescue missions at sea. This underscores the ongoing humanitarian issues related to migration across the African continent.

   • International Conferences: The images of conferences, often with political figures or international organization logos (like the United Nations), suggest that these are high-level discussions on global policies affecting Africa, such as trade, climate change, and political alliances.

   • Health and Pandemic: There are references to the COVID-19 pandemic, with images of the virus and related health measures. This indicates the impact of global health crises on African mobility and socio-economic conditions.

   • Economic Development: The presence of financial institution logos (like Ecobank) and images of economic graphs and meetings point towards economic discussions. This includes topics like financial inclusion, economic growth, and investments in Africa.

3. Narratives:

   • Human Mobility and Migration: The cluster prominently features narratives around human movement, particularly maritime migration. It highlights the challenges faced by migrants, including perilous journeys and the socio-political factors driving migration.

   • Political and Economic Governance: The frequent images of conferences and political figures suggest a narrative focused on governance, policy-making, and international relations. It highlights the importance of political stability, international cooperation, and economic policies in shaping human mobility.

   • Impact of Global Events: The references to the COVID-19 pandemic and international financial institutions indicate a narrative on how global events and economic structures impact mobility and livelihoods in Africa.

Light Colour Cluster

1. General Themes:

   • Maps and Geographic Locations: The cluster predominantly features maps of various regions in Africa. These maps include both detailed geographic data and highlighted areas, indicating specific regions of interest or concern.

   • Conflict and Crisis Zones: Many maps have red highlighted areas, suggesting regions experiencing conflict, crises, or significant events. This includes areas marked by internal displacement, refugee movements, or other humanitarian issues.

   • Routes and Paths: Several maps show routes or paths, possibly depicting migration routes, trade paths, or historical movements.

   • Regions and Territories: The maps focus on various countries and regions, showing administrative boundaries, territorial disputes, or areas of strategic importance.

2. Specific Observations:

   • Highlighted Areas: The red highlights on many maps likely indicate areas of significant activity or concern, such as conflict zones, areas of political unrest, or regions affected by natural disasters.

   • Diverse Regions: The maps cover a wide range of African regions, including North Africa, the Horn of Africa, Central Africa, and others. This suggests a comprehensive look at human mobility across the continent.

   • Contextual Information: Some maps are accompanied by logos or additional text (e.g., "WorldRemit"), indicating sources of data or organizations involved in the analysis, such as humanitarian groups, financial services, or international agencies.

   • Natural and Man-made Features: The maps include both natural features (rivers, lakes, mountains) and man-made features (borders, cities, infrastructure), providing a holistic view of the regions.

3. Narratives:

   • Human Mobility and Displacement: The red-highlighted areas and routes on the maps suggest narratives around human displacement due to conflict, economic necessity, or environmental factors. It reflects the ongoing issues of internal and cross-border migration within Africa.

   • Conflict and Crisis Management: The focus on specific conflict zones and crisis areas highlights the importance of monitoring and managing these regions. It underscores the impact of political instability and conflict on human mobility.

   • Geopolitical Analysis: The maps offer insights into the geopolitical landscape of Africa, showing areas of strategic importance, territorial disputes, and regional alliances. This narrative explores how geopolitics influence movement and migration patterns.

   • Humanitarian and Development Efforts: The presence of organizational logos and highlighted regions indicates the involvement of humanitarian and development agencies. This narrative focuses on efforts to address crises, provide aid, and support affected populations.

This cluster provides a detailed geographic and geopolitical context for understanding human mobility in Africa. If you have more clusters to analyze or need further insights, please share them.

Brown-Orange Cluster

1. General Themes:

   • Rural and Agricultural Life: The cluster includes numerous images of rural settings, agricultural activities, and natural landscapes. This suggests a focus on farming, food production, and rural communities.

   • Humanitarian Activities: There are scenes depicting aid distribution, community gatherings, and interactions with humanitarian workers. This indicates efforts to support communities facing hardship.

   • Environmental Challenges: Images of drought, dry land, and possibly desertification highlight environmental issues impacting rural areas. This includes the impact of climate change on agriculture and livelihoods.

   • Cultural and Community Events: The presence of community gatherings and cultural events points towards a focus on the social fabric and traditions of rural communities.

2. Specific Observations:

   • Agriculture and Food Production: There are images of crops, farming activities, and harvested goods. This highlights the importance of agriculture in sustaining rural communities and the challenges they face.

   • Drought and Environmental Stress: The images of dry, cracked earth and barren landscapes emphasize the environmental stress and the impact of climate change on these regions.

   • Community and Aid: Scenes of people receiving aid, community meetings, and interactions with humanitarian workers show the efforts to support these communities. This includes food distribution, medical aid, and other forms of support.

   • Cultural Practices: The images of people engaging in traditional activities and community events underscore the cultural richness and resilience of rural populations.

3. Narratives:

   • Agricultural Livelihoods: The cluster tells a story of the centrality of agriculture to rural life in Africa. It highlights both the productivity and the challenges faced by farmers, including environmental stress and climate change.

   • Humanitarian Aid and Support: The presence of aid distribution and humanitarian workers points to ongoing efforts to support vulnerable populations. This narrative emphasizes the importance of external support in times of crisis.

   • Environmental Challenges: The images of drought and barren land illustrate the severe environmental challenges impacting rural communities. This narrative calls attention to the need for sustainable practices and climate resilience.

   • Community Resilience and Culture: Despite the challenges, the images of community gatherings and cultural events highlight the resilience and vibrancy of rural African communities. This narrative celebrates the strength of social bonds and cultural traditions.

This cluster provides a deep look into the daily lives, challenges, and resilience of rural communities in Africa.

Grey Cluster

1. General Themes:

   • Urban and Industrial Scenes: Many images depict urban environments, industrial settings, and infrastructure, suggesting themes of urbanization, economic activity, and industrial development.

   • Conflict and Security: There are images of military personnel, police, and security-related activities. This points towards issues of security, law enforcement, and possibly conflict zones.

   • Economic and Financial Issues: Images of money, banking, and economic symbols highlight financial matters, economic challenges, and economic activities.

   • Migration and Refugee Situations: Scenes of people on the move, crowded places, and possibly refugee camps indicate themes of migration and displacement.

   • Political and Social Movements: There are images of political events, protests, and social gatherings, pointing to social and political movements.

2. Specific Observations:

   • Security and Law Enforcement: The presence of military and police forces suggests a focus on security measures, possibly in response to conflict or unrest. This includes both urban policing and military operations.

   • Migration and Refugee Issues: Images of people in transit, waiting in lines, and makeshift camps highlight the challenges faced by migrants and refugees. This includes the journeys they undertake and the conditions they endure.

   • Economic Activity: The images related to finance, such as stacks of money and banking activities, indicate the economic conditions impacting urban areas. This includes both formal and informal economic activities.

   • Urban Life and Infrastructure: Urban settings and infrastructure development are prominently featured, showing the growth of cities and the challenges of urbanization.

   • Political Events: Images of political leaders, protests, and polling stations suggest active political engagement and social movements within these urban environments.

3. Narratives:

   • Urbanization and Economic Activity: The cluster tells a story of the rapid urbanization taking place across Africa. It highlights the growth of cities, the economic activities that sustain them, and the infrastructure challenges they face.

   • Security and Conflict: The presence of security forces and conflict-related scenes suggests ongoing issues of safety, law enforcement, and conflict management. This narrative explores how urban areas deal with security concerns and political unrest.

   • Migration and Refugees: The images of migration and refugee situations narrate the challenges faced by displaced populations. It underscores the social and economic factors driving migration and the humanitarian response to these crises.

   • Political and Social Movements: The scenes of political events and social gatherings highlight the active engagement of urban populations in political processes. This narrative focuses on the desire for change, representation, and social justice.

General narratives from Google Cloud Vision web entities

The network below illustrates the co-occurrence of web entities related to human mobility in the African continent. The graph includes 7,419 nodes and 69,854 edges, forming an undirected graph. The ForceAtlas2 layout algorithm was applied to emphasize the structure of the network, highlighting densely connected clusters. Nodes represent unique web entities identified from images in Telegram messages, while edges indicate co-occurrence relationships between these entities. The visualization focuses on entities with a degree greater than 10 and a frequency higher than 25.

Prominent web entities in the network include "United Nations," "Human migration," "Immigration," and "Humanitarian aid." The most frequent web entity, "Refugee," appears 398 times, underscoring its centrality in discussions related to human mobility in Africa. Screenshots of Telegram messages illustrate specific instances of these entities, providing context and highlighting the breadth of topics covered. Notable messages discuss the UK's refugee policy, Uganda's acceptance of refugees, and the humanitarian crisis in Darfur.

DIscussion

The network analysis provides insights into the discourse surrounding human mobility within the investigated sample of the African Telegram sphere. The prominence of the "Refugee" web entity, identified by the Cloud Vision API, as the primary label for the collection of images analyzed, reflects the global attention and concern regarding African refugee issues.

Entities such as "United Nations" and "Humanitarian aid" frequently appear in the images, indicating that the discourse is often framed within international and humanitarian contexts. This aligns with the global and collaborative efforts required to tackle human mobility challenges. Such entities suggest that the images capture themes related to aid and support provided by international organizations, emphasizing the importance of global cooperation in addressing the complexities of human migration.

The presence of the "Climate change" web entity suggests connection between climate-related events and human mobility in Africa. This highlights the impact of climate-induced displacement, conflicts, and migrant boat crossings on the movement of people, whether voluntary or forced.

The graph also reveals thematic clusters, with entities associated with specific regions (e.g., "Sudan," "Rwanda") and themes (e.g., "Human migration," "Immigration policy"). These clusters offer insights into regional variations in discourse and the diverse factors influencing human mobility across the continent. They can provide a layer of understanding on how different regions are affected by and respond to mobility-related challenges.

In summary, this network visualization offers a snapshot of prevalent themes and potential areas for further research aimed at understanding the complex narratives around human mobility shared on Telegram. However, it's essential to consider the interpretation limitations of the Google Cloud Vision API, as it may not fully capture the context and subjective perspectives that shape these narratives.

The two artificial intelligence tools employed in interpreting the images – GPT 4.0 and Google Cloud Vision API – play a crucial role in reflecting human mobility visual vernaculars. However, these tools have limitations. Without context about the selected groups and channels, the AI-generated results may lack the subjective perspectives and nuances that drive the narratives and framings of these images. We found that GPT was more effective for analyzing color clusters, providing deeper insights into visual patterns. On the other hand, the Google Cloud Vision API was more suited for processing the entire set of images, offering a broader view of the web entities present.

e) Afghanistan and Iran

Researchers’ motivation:I am motivated to join the research project: “uncovering visual narratives about human mobility on Telegram” due to my extensive experience working with Afghan women refugees and children in Iran for more than 12 years. This background has provided me with a deep understanding of the challenges and resilience of Afghan immigrants in Iran. Participating in this (case)study allows me to compare the visual narratives discovered on social media with my qualitative experiences and observations in the field. I am particularly interested in how these narratives differ or align with the lived experiences of the communities I have worked with. This comparative analysis will offer valuable insights and a more nuanced understanding of the Afghan immigrant experience. Moreover, I am eager to explore the role of social media in shaping and disseminating these narratives. By focusing on Telegram, I aim to understand how visual content on this platform influences public perception and the self-representation of Afghan immigrants. This aspect of the study will contribute to a broader understanding of the intersection between social media and migration narratives. Additionally, I am keen to compare our findings with those related to other immigrant groups. This comparative approach will help identify common themes and unique challenges across different migrant communities, providing a comprehensive perspective on human mobility narratives. I am also interested in sharing the results of our study with Afghan refugees and learning about their perspectives on the findings. Engaging with the refugee community to discuss these digital narratives will provide valuable feedback and help validate our research. Furthermore, I aim to use the analytical tools in an engaging way to explore digital narratives alongside the immigrants themselves, fostering a collaborative and inclusive research process.

Telegram groups and channels list curation and TeleCatch keyword querying

• I started with a list of Telegram channels and groups relevant to Afghan immigrant communities, drawing on my familiarity with existing Afghan-focused channels. This initial list included channels where Afghan refugees actively engage and discuss their experiences.To expand this list I performed the following procedures:

  • Snowball Sampling: I reviewed the memberships and related channels of these initial Afghan-focused groups.

  • Similar Channels: I used Telegram’s “Similar Channels” feature to identify additional relevant channels and groups.

  • Recommendations and Mentions: I explored recommendations and mentions within these groups to discover new channels.

• Keyword Translation and Querying: To ensure comprehensive coverage, I translated relevant keywords from the taxonomy related to Afghan migration, refugee experiences, and related issues into Persian. In some cases, I queried these terms in both their translated forms and in English, to capture both the local and international discourse surrounding Afghan migration.

• After uploading the curated list of groups and channels to TeleCatch, I qualitatively reviewed their content to ensure they were relevant to Afghan immigrant narratives. Then, I queried their content with the keywords, including both the translated terms and English terms, to capture a wide range of content.

Main findings

Chat GPT interpretations

Observations from entire image wall

1. Diverse Contexts and Situations:

• The images seem to depict a wide range of situations and contexts. This includes daily life activities, migration journeys, interactions with aid organisations, and different geographical settings

2. Humanitarian Assistance:

• There are numerous images that likely show the involvement of humanitarian aid—food distribution, medical assistance, shelters, and support from various organisations. These images highlight the role of local and international efforts in supporting refugees and immigrants.

3. Migration Routes and Transportation:

• Some images appear to capture the journey itself, showing transportation modes like buses, trucks, and even people walking with their belongings. This highlights the arduous journeys undertaken by Afghan refugees

4. Living Conditions:

• Various images illustrate the living conditions in refugee camps or temporary settlements. This includes makeshift shelters, tents, and communal living spaces that reflect the temporary and often challenging living situations of refugees.

5. Cultural and Social Life:

• There are glimpses of cultural and social activities—children playing, families gathering, traditional dress, and cultural expressions. These images convey a sense of community and resilience despite difficult circumstances.

6. Challenges and Hardships:

• The presence of images showing hardship, such as crowded conditions, long queues for aid, and expressions of distress, underscores the challenges faced by Afghan refugees and immigrants.

7. Integration and Adaptation:

• Some images likely depict efforts toward integration and adaptation in new environments, such as language classes, job training, and interaction with local communities

8. Protests and Advocacy:

• There may be images related to protests, advocacy, and awareness campaigns, highlighting the political and social dimensions of refugee and immigrant issues.

Observations from colour clusters

Black and grey cluster

1. Text-Based Content:

   • There are numerous images with black backgrounds and white or coloured text. These seem to include quotes, statements, and possibly tweets or social media posts. They could be conveying information, opinions, and news updates about the situation of Afghan refugees and immigrants.

   • Some of these text-based images appear to include official announcements, personal testimonies, or advocacy messages.

2. Public Figures and Leaders:

   • Many images feature public figures, politicians, and possibly religious leaders. These images likely represent speeches, interviews, or statements made by influential individuals about the refugee crisis.

   • The presence of various leaders suggests discussions around policies, international relations, and decisions impacting Afghan refugees.

3. Media and News Footage:

   • There are images that seem to be snapshots from news broadcasts or media coverage. These include images of protests, political gatherings, and events related to the refugee situation.

   • Some of these images might be from news reports, showing the gravity of the situation, key developments, or highlighting significant incidents.

4. Emotive and Impactful Imagery:

   • Several images appear to show emotional or impactful scenes. This includes close-up shots of people in distress, scenes of conflict, and other evocative photographs that convey the human side of the refugee crisis.

   • These images are likely intended to evoke empathy and draw attention to the humanitarian aspects of the crisis.

5. Social Media Elements:

   • Some images include recognizable elements from social media platforms, such as twitter15 layouts and social media post formats. This indicates the role of social media in disseminating information, raising awareness, and mobilising support for Afghan refugees.

6. Diverse Contexts and Themes:

   • The cluster covers a broad range of themes, from political and social commentary to personal stories and official communications. This diversity reflects the complexity of the refugee issue and the various dimensions involved.

The black colour cluster from the mosaic predominantly features text-based content, public figures, media coverage, and emotive imagery, encapsulating a wide array of narratives related to Afghan refugees and immigrants. This includes official statements, social media discourse, political commentary, and human interest stories.

People cluster

1. Migration and Mobility:

• Several images show groups of people traveling, often with luggage or belongings, indicating migration or movement. This suggests ongoing mobility among the refugee population, possibly due to seeking better living conditions or escaping conflict.

2. Crowded Conditions:

• Many photos depict large crowds, either in refugee camps or urban settings. This highlights the high population density in areas where Afghan refugees are concentrated, which could be indicative of overcrowded living conditions.

3. Living Conditions:

• There are images showing makeshift shelters and camps, which suggest that many refugees live in temporary or substandard housing. This can be indicative of the challenging living conditions faced by refugees.

4. Aid and Assistance:

• Several photos display the distribution of aid, such as food, medical care, and other forms of assistance. This underscores the ongoing humanitarian efforts to support Afghan refugees in Iran.

5. Military and Security Presence:

• Some images include military personnel or checkpoints, suggesting a significant security presence. This could imply that the mobility and activities of refugees are closely monitored.

6. Community and Social Interactions:

• Group photos and gatherings indicate the presence of community activities and social interactions among refugees. These interactions can be crucial for maintaining social cohesion and support networks.

7. Health and Medical Care:

• A few images show medical personnel attending to refugees, highlighting health issues and the need for medical care within the refugee population.

8. Protests and Demonstrations:

• Some images appear to depict protests or demonstrations, indicating that refugees may be actively voicing their grievances or demands

9. Children and Families:

• Numerous images feature children and families, emphasising that a significant portion of the refugee population includes families with young children.

10. Economic Activities:

• There are photos showing markets and economic activities, suggesting that refugees are engaged in various forms of livelihood and trade to sustain themselves.

11. Education:

• Some images depict educational settings or activities, highlighting efforts to provide education to refugee children.

Orange cluster

1. Humanitarian Aid and Relief Efforts:

• Distribution Events: Scenes of aid distribution centers or relief efforts with an orange hue might indicate organized humanitarian activities aimed at providing essential services to refugees.

• Emergency Responses: Photos of emergency situations, such as fires or rescue operations, where orange tones are prominent, highlight urgent humanitarian responses.

2. Children and Education:

• School Activities: Orange-coloured classrooms, school supplies, or uniforms indicate the focus on education for refugee children, showcasing efforts to provide learning opportunities in challenging circumstances.

• Play and Recreation: Images of children playing in environments with orange elements reflect the importance of normalcy and joy amidst the difficult conditions of displacement.

3. Protest and Advocacy:

• Vibrant Demonstrations: Photos of protests or advocacy events with banners, posters, or clothing in orange signify the active efforts of refugees and supporters to demand rights and raise awareness.

• Solidarity Movements: Visuals of solidarity events or campaigns using orange as a theme colour underscore the collective action taken by communities and allies.

4. Political and Social Commentary:

• Orange-themed backgrounds in images of speeches, press conferences, or public statements by political figures may indicate specific political events or announcements related to refugee policies.

• Screenshots from media coverage using orange tones could highlight key stories or impactful events covered by journalists.

Discussion

Hierarchized Publics and Dominant Discourses

On Telegram, discussions about human mobility, particularly Afghan migration, often intersect with dominant discourses, such as broader political or geopolitical narratives. This intersection can shape the visibility and interpretation of specific issues. Afghan immigrant experiences are frequently framed within these larger contexts, which can obscure their unique aspects and lead to their marginalisation.

Mechanisms of Inclusion and Exclusion

1. Framing Through Dominant Discourses: The representation of Afghan migration is frequently aligned with overarching themes like regional conflict or political instability. This framing can limit the focus on specific refugee experiences and challenges.

2. Participatory Dynamics: Engaging Afghan immigrants in curating channels and keywords ensures that their narratives are better represented. This participatory approach helps mitigate issues of exclusion by amplifying the voices of those directly affected.

Human Mobility Representation: AI Interpretation of Visual Vernaculars

1. Effective Pattern Recognition: AI tools effectively identify visual patterns and themes related to human mobility, providing valuable insights into underlying narratives.

2. Limitations and Caveats: While AI models and ChatGPT offer useful analyses, they should be supplemented with qualitative reviews to avoid vague or generalised interpretations.

3. Future Research Opportunities: Network analysis of AI-generated descriptions presents promising avenues for further research into visual narratives and their implications.

f) Ukraine and Visegrad

Researchers’ motivation: As an individual who has been affected by the ongoing Ukraine - Russia conflict, having lived most of my life in Slovakia, which borders Ukraine, I was keen on joining this particular project about human mobility on Telegram. Through this conflict, Slovakia has received a substantial number of Ukrainian refugees fleeing the war, which, due to the fractured nature of communication in chaotic times, mobilise primarily through online outlets such as social media. As Telegram paved its way and now serves as one of the primary information channels around the Russia - Ukraine war, I was keen on analysing it through this lens. What I am particularly interested in exploring, is how social media affects how people mobilise, and how they can even shape narratives around the conflict.

Telegram groups and channels list curation and TeleCatch keyword querying

As a native speaker of Slovak, I focused on Telegram channels that were in Slovak, later on this channel pool was expanded to include Ukrainian as well as any stray languages coming from the Visegrad group of countries. Before using the TeleCatch tool, I compiled a list of relevant channels in the following manner: I translated most of the keywords the team agreed upon would be used in their respective channel collection, using these translated queries yielded a couple of relevant channels at first. To expand the list even further, the Telegram “Similar channel” feature was used, where Telegram recommends the user similar channels and groups to the one they are currently on. Then all of these curated channels and groups were run through the TeleCatch tool to get qualitative data in the form of images and more, vital for further analysis.

11 Groups, 507 images, languages are - Slovak and Ukrainian

Main findings

Chat GPT interpretations

Observations from entire image wall

1. Humanitarian Efforts and Aid:

   • There is a significant presence of images from organizations like the UNHCR, with logos, infographics, and documents that highlight efforts in providing aid to displaced persons.

   • Various slides and information pamphlets provide details on aid distribution, support services, and relief programs.

2. Political and Diplomatic Activities:

   • Images of political leaders, press conferences, and diplomatic meetings suggest ongoing international discussions and negotiations regarding the conflict.

   • This cluster includes both local and international political figures engaging in dialogue or addressing the media.

3. Military and Conflict Scenes:

   • Numerous images depict scenes from the battlefield, including military personnel, equipment, and action shots of conflicts.

   • Photos of destruction, damaged infrastructure, and the aftermath of battles highlight the intense impact of the conflict on local areas.

4. Public and Social Reactions:

   • Protests, large gatherings, and public demonstrations show the societal response to the conflict.

   • Images of people holding banners, participating in rallies, and engaging in community activities emphasize collective action and advocacy.

5. Maps and Geographic Information:

   • Detailed maps and geographic data indicate the movements of displaced populations, strategic locations, and areas affected by the conflict.

   • These visuals provide a spatial understanding of the conflict's impact on human mobility and territorial changes.

6. Media Coverage and Interviews:

   • Still shots from interviews, news anchors, and broadcast segments reflect the media's role in covering the conflict and disseminating information.

   • The presence of various news channels and reporting styles suggest a diverse media landscape documenting the crisis.

Thematic Clusters

Blue Cluster (Top Center)

• Humanitarian aid organizations' logos, such as UNHCR.

• Informational graphics and resources about support services, including shelter, food, and medical assistance.

• Maps and charts showing refugee movements and aid distribution.

Yellow Cluster (Middle Right)

• Information slides and guides, possibly from educational or public awareness campaigns.

• Visual aids detailing health and safety measures, community support programs, and emergency response initiatives.

White Cluster (Center)

• Documents, reports, and forms that appear to be official publications or administrative records.

• Likely includes guidelines, policies, and statistical data relevant to the conflict and humanitarian efforts.

Grey and Dark Clusters (Top Right)

• Images related to military operations, conflict zones, and on-ground action.

• Photographs of destruction, affected civilian areas, and military personnel in action.

• Coverage of political figures, press conferences, and strategic discussions.

Colourful Cluster (Bottom Left and Center)

• Family and community images depicting everyday life amidst the conflict.

• Scenes of people engaging in cultural activities, social events, and personal stories of resilience.

• Emphasizes the human aspect and personal experiences during the conflict.

Web entity co-occurrence graph making process

Filter: Topology (Degree Range): 13-385

Layout: ForceAtlas2

Appearance: Nodes: Ranking:

• Filter: Topology (Degree Range): 13-385

• Layout: ForceAtlas2

• Appearance: Nodes: Ranking:

• Size: 8-50

• Color Scale:

Discussion

When we analysed the image and keyword clusters, two trends arose, which were picked up by the AI algorithm as well. Firstly, we can observe the two most dominant clusters, which are the yellow and blue clusters, followed by the white cluster. These are symbolic of the Ukrainian national colours and follow suit in their sentiment, usually being informational and helpful in nature, in form of short tips or updates on the life of Ukrainian migrants in Slovakia trying to get by.

The second trend that arose are the colourful and grey clusters, which AI labelled as images related to military operations, conflict zones, and on-ground action. These serve as a more raw representation of the ongoing conflict, not shying away from even gore, which seems to be a trend on Telegram, as was expected from a social media known for its lax attitude on moderation and what content is allowed on their websites. From images alone it is hard to ascertain whether these pictures play an informative or shocking and fearful role, so it is safe to assume it is true for both. These can be used by more trusted channels to inform on the ongoing conflict, or by more populist channels for fear mongering, which were definitely present and exist, due to the ongoing split opinion the Slovak general public has on the conflict in Ukraine.

The last section continues in the same aforementioned argument, as the colourful cluster, though mislabeled by the AI for cultural and community images, upon thorough view, depicts highly politicised material featuring the likes of Trump, Lukashenko and even Wilders, amongst the more expected results of people such as Zelenskyy. Although AI seemed to hold its own in describing almost every image cluster, it is safe to pinpoint two shortcomings that have been observed in this theme as well as the others. A preference on overgeneralization, which may be caused by the failings of the algorithm system or perhaps due to the controversial nature of the subject matter, is one of its main shortcomings. As well as the obvious mislabeling, which means there is always a need for human cross examination when utilising ChatGPT or any other AI systems for analysis.

6. Conclusions

Considering the pertinence of human mobility to contemporary social dynamics, this research project explored the potential of visual analysis to identify framings, symbols and values from imagery that circulate on digital platforms. The presence of such narratives were found across Telegram groups and channels not explicitly focused on the topic. Related debates were present in both specific interest groups, such as nationalists, conspiratorial, and climate action advocates, as well as in forums dedicated to geolocated topics. The images shared within these groups, extracted through queries specific to human mobility, encapsulate broader visual representations of migration and displacement, with their subjective meanings dependent on the context of their respective world visions and/or geographic coverage.

In examining the effects of right-wing extremism on migration perceptions in the Global North, the study analysed perspectives from conspiratorial and nationalist propagandist viewpoints, juxtaposing these with counter-narratives from climate activist communities. Within conspiracy-focused groups and channels, recurring themes included the surveillance state, immigration and border controls, geopolitics, and political figures. Nationalist groups addressed similar issues, such as large-scale migration, law enforcement and border control, detention facilities, and political figures. Conversely, discussions about human mobility were largely absent from climate action groups and channels. Nevertheless, limited visual content related to human mobility in these groups depicted climate-induced migration, solidarity for refugees, and broader impacts of climate change.

The analysis also focused on three geographical regions with distinct experiences of human mobility: the African continent, the movement of Ukrainian refugees into the Visegrad group countries (Czech Republic, Hungary, Poland, and Slovakia), and the migration of Afghanistani people into Iran. The African dataset revealed images reflecting climate-induced displacement, conflict, and migrant boat crossings. The dataset concerning Ukrainian refugees in Slovakia included representations of the Russia-Ukraine war, humanitarian institutions, refugee support activities, and political figures. In the Afghanistani dataset, images depicted daily life activities of refugees, migration journeys, interactions with aid organisations, and engagement with Iranian social issues.

From a methodological perspective, we explored the ability of Artificial Intelligence to interpret these images and effectively compose a visual vernacular (or vernaculars) of human mobility representations. The AI tools employed to interpret the collected images—GPT 3.0 and 4.0, as well as Google Cloud Vision API—effectively captured the visual vernaculars of human mobility. However, without providing context about the selected groups and channels, the AI's results lacked the subjective interpretation that drive these narratives and framings.

From a thematic perspective, this study enhances the understanding of public perceptions, concerns and priorities in relation to human mobility as represented in a public dialogue space such as Telegram. While human mobility is visually signified according to the specific, subjective values that drive each of our case studies, the application of AI enables distinguishing common threads across large datasets and the identification of emerging trends.

7. References

Alexander, J. C., & Smith, P. (2010). The strong program: Origins, achievements, and prospects. Handbook of cultural sociology, 13-24. https://ccs.yale.edu/about-us/strong-program-origins-achievements-and-prospects

Bastian, M., Heymann, S., & Jacomy, M. (2009). Gephi: An open source software for exploring and manipulating networks. [Computer software]. https://gephi.org/

Carneiro, B., Pacillo, G., Läderach, P. R. D., DuttaGupta, T., & Resce, G. (2022). Mapping Policymaker Perspectives Of The Climate Security- Migration Nexus In Nigeria: A Social Media Analysis. https://hdl.handle.net/10568/127619

Carneiro, B., & Tucci, G. (2024). Using TeleCatch to investigate human mobility discourses in the African Telegram sphere. https://hdl.handle.net/10568/145322

Carneiro, B., Tucci, G., Gomes, F., Cruciata, P., & Resce, G. (2023a). Mapping public debates about the climate security-human mobility nexus in Jordan: A social media analysis. https://hdl.handle.net/10568/135950

Carneiro, B., Tucci, G., Gomes, F., Cruciata, P., & Resce, G. (2023b). Mapping public debates about the climate security-mobility nexus in Mozambique: A social media analysis. https://hdl.handle.net/10568/136035

Chao, J. (2023). Memespector-GUI: Graphical User Interface Client for Computer Vision APIs (Version 0.2.5) [C#]. https://doi.org/10.5281/zenodo.7704877

Cochrane, C., & Nevitte, N. (2014). Scapegoating: Unemployment, far-right parties and anti-immigrant sentiment. Comparative European Politics, 12, 1-32. https://link.springer.com/content/pdf/10.1057/cep.2012.28.pdf

Colombo, G., Bounegru, L., & Gray, J. (2023). Visual Models for Social Media Image Analysis: Groupings, Engagement, Trends, and Rankings. International Journal of Communication, 17(0), Article 0.

Ekman, M. (2022). The great replacement: Strategic mainstreaming of far-right conspiracy claims. Convergence, 28(4), 1127-1143. https://journals.sagepub.com/doi/pdf/10.1177/13548565221091983

Gibbs, M., Meese, J., Arnold, M., Nansen, B., & Carter, M. (2015). #Funeral and Instagram: Death, social media, and platform vernacular. Information, Communication & Society, 18(3), 255–268. https://doi.org/10.1080/1369118X.2014.987152

Google Cloud. (n.d.). Vision AI | Cloud Vision API. Google Cloud. Retrieved May 15, 2023, from https://cloud.google.com/vision

IOM. (2022). Interactive World Migration Report 2022. https://worldmigrationreport.iom.int/wmr-2022-interactive/

Kaufmann, E. (2017). Why values, not economics, hold the key to the populist right-and to crafting new migration narratives. Migration Research Leaders’ Syndicate, 183. https://publications.iom.int/es/system/files/pdf/migration_research_leaders_syndicate.pdf#page=191

Kosec, K., Läderach, P., & Ruckstuhl, S. (2023, May 22). Fragility, Conflict, and Migration: A new CGIAR Initiative for a fragile world. CGIAR. https://www.cgiar.org/news-events/news/fragility-conflict-and-migration-a-new-cgiar-initiative-for-a-fragile-world/

McLeman, R. (2019). Climate and Human Migration: Past Experiences, Future Challenges. Cambridge University Press.

Mitts, T. (2022). Countering Violent Extremism and Radical Rhetoric. International Organization, 76(1), 251–272. https://doi.org/10.1017/S0020818321000242

Mouffe, C. (2011). On the political. Routledge. https://www.taylorfrancis.com/books/mono/10.4324/9780203870112/political-chantal-mouffe

OpenAI. (2023). ChatGPT [Computer software]. https://openai.com/chatgpt/

Rogers, R. (2020). Deplatforming: Following extreme Internet celebrities to Telegram and alternative social media. European Journal of Communication, 35(3), 213–229. https://doi.org/10.1177/0267323120922066

Ruscica, G., Tucci, G., & Carneiro, B. (2023, August 6). TeleCatch. GitLab. https://gitlab.com/LaBaffa/teledash

Savelli, A., Schapendonk, F., Dutta Gupta, T., Pacillo, G., & Läderach, P. (2023). Climate change, mobility and violent conflict: A typology of interlinked pathways. International Development Planning Review, 45(4), 403–436. https://doi.org/10.3828/idpr.2023.2

Tucci, G. (2023). Além da mensagem: O Telegram como um ambiente informacional complexo, suas affordances e a desinformação política no Brasil [Universidade Federal do Rio de Janeiro]. http://dx.doi.org/10.13140/RG.2.2.10362.91842

UNHCR. (2024). Global Trends report 2023. UNHCR. https://www.unhcr.org/global-trends-report-2023

Visual Computing Group. (2018). ImageSorter [Computer software]. HTW Berlin. https://visual-computing.com/project/imagesorter/

Yılmaz, F. (2012). Right-wing hegemony and immigration: How the populist far-right achieved hegemony through the immigration debate in Europe. Current sociology, 60(3), 368-381. https://journals.sagepub.com/doi/pdf/10.1177/0011392111426192

8. Acknowledgements

This work was carried out with support from the CGIAR Initiative on Climate Resilience (ClimBeR), and the CGIAR Initiative on Fragility, Conflict, and Migration (FCM). We would like to thank all funders who supported this research through their contributions to the CGIAR Trust Fund: https://www.cgiar.org/funders/.