Christopher Starke
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Christopher Starke has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
- Nils Köbis
Research Center Trustworthy Data Science and Security, University of Duisburg-Essen, Duisburg, Germany
6 shared publications Claes de Vreese
5 shared publications- Marco Lünich
Department of Social Sciences, Heinrich Heine University, Düsseldorf, Germany
4 shared publications Natali Helberger
4 shared publications- Alfio Ventura
Research Center Trustworthy Data Science and Security, University Alliance Ruhr
3 shared publications - Frank Marcinkowski
Heinrich Heine University Düsseldorf, Germany
3 shared publications
Latest Publications
- Open Access
- 7 October 2026
- Published by Center for Open Science in SocArXiv
Conceptualizing Political Consumerism in the Attention Economy
Social media platforms increasingly become targets of politically and ethically motivated user protest, yet existing approaches do not sufficiently explain how deliberate platform use can operate as political participation. This conceptual paper extends the established concept of political consumerism to the attention economy, defining it as individuals’ market-oriented behaviors to punish or reward social media platforms on political or ethical grounds. It argues that three main factors constrain political consumerism on social media platforms: opaque value extraction, limited ethical alternatives, and high switching costs. Because political consumerism can take many forms, the paper suggests distinguishing user actions by their economic impact and whether they signal user preferences to the platform company. It further explains how social media literacy and regulation can improve the legibility and actionability of these practices. Finally, it outlines methodological implications for studying political consumerism through mixed-method designs that combine user motivations with behavioral data.
- 7 October 2026
- Published by Center for Open Science in SocArXiv
Open AccessConceptualizing Political Consumerism in the Attention Economy
Social media platforms increasingly become targets of politically and ethically motivated user protest, yet existing approaches do not sufficiently explain how deliberate platform use can operate as political participation. This conceptual paper extends the established concept of political consumerism to the attention economy, defining it as individuals’ market-oriented behaviors to punish or reward social media platforms on political or ethical grounds. It argues that three main factors constrain political consumerism on social media platforms: opaque value extraction, limited ethical alternatives, and high switching costs. Because political consumerism can take many forms, the paper suggests distinguishing user actions by their economic impact and whether they signal user preferences to the platform company. It further explains how social media literacy and regulation can improve the legibility and actionability of these practices. Finally, it outlines methodological implications for studying political consumerism through mixed-method designs that combine user motivations with behavioral data.
- Open Access
- 15 September 2026
- Published by Center for Open Science in SocArXiv
Able but unwilling? Mapping citizens’ political GenAI use through the lens of Expectancy-Value Theory
Generative AI (GenAI) holds promise and risk for democracy, yet little is known about howcitizens evaluate and adopt it for political purposes. Drawing on Expectancy-Value Theoryand Uses & Gratifications Theory, this study offers an initial account of political GenAIadoption, mapping how citizens' cost-value perceptions and expectancy beliefs cluster intodistinct profiles. We investigate how these ‘expectancy-value profiles’ shape political GenAIuse. Using cross-sectional survey data from 947 Dutch respondents, we conduct a latentprofile analysis, identifying four profiles: optimists, cautious pragmatists, sceptics, andindifferents. Profile differences stem largely from differing cost-value perceptions rather thanexpectancy beliefs, suggesting the question is less about whether people feel able to useGenAI, and more about whether they feel it is worth using. Across all profiles, costperceptions are high and political use is low, though optimists report significantly more usethan others. Individuals’ political dispositions predict usage but not profile membership andamong technological dispositions, trust in GenAI and belief in the machine heuristic predictbeing an optimist. Findings raise concerns about uncalibrated trust among adopters.
- 15 September 2026
- Published by Center for Open Science in SocArXiv
Open AccessAble but unwilling? Mapping citizens’ political GenAI use through the lens of Expectancy-Value Theory
Generative AI (GenAI) holds promise and risk for democracy, yet little is known about howcitizens evaluate and adopt it for political purposes. Drawing on Expectancy-Value Theoryand Uses & Gratifications Theory, this study offers an initial account of political GenAIadoption, mapping how citizens' cost-value perceptions and expectancy beliefs cluster intodistinct profiles. We investigate how these ‘expectancy-value profiles’ shape political GenAIuse. Using cross-sectional survey data from 947 Dutch respondents, we conduct a latentprofile analysis, identifying four profiles: optimists, cautious pragmatists, sceptics, andindifferents. Profile differences stem largely from differing cost-value perceptions rather thanexpectancy beliefs, suggesting the question is less about whether people feel able to useGenAI, and more about whether they feel it is worth using. Across all profiles, costperceptions are high and political use is low, though optimists report significantly more usethan others. Individuals’ political dispositions predict usage but not profile membership andamong technological dispositions, trust in GenAI and belief in the machine heuristic predictbeing an optimist. Findings raise concerns about uncalibrated trust among adopters.
- Open Access
- 25 June 2026
- Published by Elsevier in Computers in Human Behavior Reports
- Vol. 23
Relationships in the age of AI: A review on the opportunities and risks of synthetic relationships to reduce loneliness
- 25 June 2026
- Published by Elsevier in Computers in Human Behavior Reports
- Vol. 23
Open AccessRelationships in the age of AI: A review on the opportunities and risks of synthetic relationships to reduce loneliness
- 9 June 2026
- Published by Edward Elgar Publishing
- p. 283-298
The public legitimacy of artificial intelligence governance
This chapter examines the public legitimacy of artificial intelligence governance (AIG) by synthesising the existing comparative, survey-based literature. We conceptualise legitimacy sociologically, encompassing the input, throughput, and output dimensions of AI-affected democratic decision-making. Reviewing evidence across countries and use cases, we show that legitimacy perceptions hinge on perceived transparency, fairness, accountability, performance, and distributive outcomes, yet vary by culture, knowledge, values, and trust in institutions and AI. Publics remain ambivalent. Awareness is uneven, politicisation is nascent, and experts and affected groups often diverge from the general public. Communication environments and participation formats shape attitudes, but superficial public involvement risks backlash. We map four research frontiers: addressing Euro-Atlantic bias by integrating Global South perspectives; decomposing layered trust across governments, companies, scientific bodies, and AI; tracing salience and politicisation and their shifts from output to input legitimacy criteria; and examining agentic systems and synthetic relationships that may reconfigure accountability, responsibility, and oversight.
- 9 June 2026
- Published by Edward Elgar Publishing
- p. 283-298
The public legitimacy of artificial intelligence governance
This chapter examines the public legitimacy of artificial intelligence governance (AIG) by synthesising the existing comparative, survey-based literature. We conceptualise legitimacy sociologically, encompassing the input, throughput, and output dimensions of AI-affected democratic decision-making. Reviewing evidence across countries and use cases, we show that legitimacy perceptions hinge on perceived transparency, fairness, accountability, performance, and distributive outcomes, yet vary by culture, knowledge, values, and trust in institutions and AI. Publics remain ambivalent. Awareness is uneven, politicisation is nascent, and experts and affected groups often diverge from the general public. Communication environments and participation formats shape attitudes, but superficial public involvement risks backlash. We map four research frontiers: addressing Euro-Atlantic bias by integrating Global South perspectives; decomposing layered trust across governments, companies, scientific bodies, and AI; tracing salience and politicisation and their shifts from output to input legitimacy criteria; and examining agentic systems and synthetic relationships that may reconfigure accountability, responsibility, and oversight.
- Open Access
- 12 May 2026
- Published by Center for Open Science in SocArXiv
surveychat: An Open-Source Platform for LLM-Based Conversational Surveys and Experiments
Surveychat is an open-source web application that enables researchers to administer surveys and conduct randomized experiments involving large language model (LLM)-based conversational agents, without the need to develop custom web application code. The system supports two primary operational modes: (i) survey mode, in which all participants interact with an identical chatbot configuration, and (ii) experiment mode, in which participants are randomly assigned to one of multiple chatbot conditions, each defined by a researcher-specified persona and language model. Upon completion of the interaction, participants receive an anonymized JSON transcript that contains only the role, content, and timestamp of each message. This transcript can be copied back into the parent survey platform (such as Qualtrics), within which the chatbot interface itself can also be directly embedded. The frontend of surveychat is implemented using Streamlit, and the entire application is configured via a single Python file. The system does not persist conversation data on its server and is compatible with any chat-completions-compatible API endpoint - including locally hosted models - thereby allowing researchers to retain full control over model selection, API usage, data jurisdiction, and adherence to ethical and regulatory requirements.
- 12 May 2026
- Published by Center for Open Science in SocArXiv
Open Accesssurveychat: An Open-Source Platform for LLM-Based Conversational Surveys and Experiments
Surveychat is an open-source web application that enables researchers to administer surveys and conduct randomized experiments involving large language model (LLM)-based conversational agents, without the need to develop custom web application code. The system supports two primary operational modes: (i) survey mode, in which all participants interact with an identical chatbot configuration, and (ii) experiment mode, in which participants are randomly assigned to one of multiple chatbot conditions, each defined by a researcher-specified persona and language model. Upon completion of the interaction, participants receive an anonymized JSON transcript that contains only the role, content, and timestamp of each message. This transcript can be copied back into the parent survey platform (such as Qualtrics), within which the chatbot interface itself can also be directly embedded. The frontend of surveychat is implemented using Streamlit, and the entire application is configured via a single Python file. The system does not persist conversation data on its server and is compatible with any chat-completions-compatible API endpoint - including locally hosted models - thereby allowing researchers to retain full control over model selection, API usage, data jurisdiction, and adherence to ethical and regulatory requirements.
- Open Access
- 2 April 2026
- Published by Elsevier in Technology in Society
- Vol. 87
With AI through the energy crisis? Citizens’ legitimacy perceptions of human and AI-supported hybrid decisions in an energy policy decision-making process
- 2 April 2026
- Published by Elsevier in Technology in Society
- Vol. 87
Open AccessWith AI through the energy crisis? Citizens’ legitimacy perceptions of human and AI-supported hybrid decisions in an energy policy decision-making process
- Open Access
- 17 November 2025
- Published by Center for Open Science in PsyArXiv
Human-in-the-Loop Oversight of AI is Compromised by Political Preferences
Humans are legally responsible for monitoring sensitive Artificial Intelligence (AI) systems, requiring a human-in-the-loop (HITL) to have the final say in approving or rejecting AI decisions. However, behavioral science shows humans are not error and bias free. We report the results of three experiments (n = 5798, 115960 decisions) modeled on real-world welfare allocation scenarios in which locals and immigrants are considered for financial support. Results revealed that HITL oversight does not reliably correct algorithmic errors, and may even exacerbate errors. Moreover, political preferences of the HITL shape the type of errors they make. Interventions incentivizing accuracy or clarifying instructions reduced HITL error rates. Nevertheless, HITL error rates were higher than algorithmic error rates. These results point out that HITL is not a silver-bullet solution across settings. Our work calls for more systematic research into when, and how, oversight of AI can effectively reduce error and improve decision making quality.
- 17 November 2025
- Published by Center for Open Science in PsyArXiv
Open AccessHuman-in-the-Loop Oversight of AI is Compromised by Political Preferences
Humans are legally responsible for monitoring sensitive Artificial Intelligence (AI) systems, requiring a human-in-the-loop (HITL) to have the final say in approving or rejecting AI decisions. However, behavioral science shows humans are not error and bias free. We report the results of three experiments (n = 5798, 115960 decisions) modeled on real-world welfare allocation scenarios in which locals and immigrants are considered for financial support. Results revealed that HITL oversight does not reliably correct algorithmic errors, and may even exacerbate errors. Moreover, political preferences of the HITL shape the type of errors they make. Interventions incentivizing accuracy or clarifying instructions reduced HITL error rates. Nevertheless, HITL error rates were higher than algorithmic error rates. These results point out that HITL is not a silver-bullet solution across settings. Our work calls for more systematic research into when, and how, oversight of AI can effectively reduce error and improve decision making quality.
- Open Access
- 31 October 2025
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 27, e63909
Themes, Policies, and Attention Shifts Regarding COVID-19 Vaccinations in German-Speaking Regions: Infoveillance Study Using Tweets
Background: Societies worldwide have witnessed growing rifts separating advocates and opponents of vaccinations and other COVID-19 countermeasures. With the rollout of vaccination campaigns, the European German-speaking region (Germany, Austria, and Switzerland) initially exhibited a noticeably low vaccination uptake compared to other European regions. Later, uptake increased. It remains unclear which factors contributed to these changes. Objective: This study aimed to shed light on the intricacies of vaccine hesitancy among the German-speaking population and the possible dynamics between policy changes and public concerns using web discourse data. These insights are valuable for policymakers tasked with making far-reaching decisions—policies need to effectively curb the spread of the virus and at the same time respect fundamental civil liberties and minimize undesired consequences. Methods: This study drew on data from Twitter (subsequently rebranded X). We used a hybrid pipeline to detect and analyze 191,750 German-language vaccination-related tweets using a semiautomatic seed list generation approach, topic modeling, sentiment analysis, and a minimum of social scientific domain knowledge to evaluate the discourse about vaccinations in light of the COVID-19 pandemic. We further analyzed the evolution of public attention during different phases of the pandemic and in relation to policy changes to identify potential drivers of shifts in public attention. Results: The discourse concerning vaccinations was associated with more negative sentiments than the general discourse on German-speaking Twitter (47,159/191,750, 24.59% vs 1,758,776/12,297,163, 14.3% predominantly negative tweets, respectively). The relative frequencies of the discussed themes fluctuated heavily (eg, safety and side effects was the most dominant theme in wave 3 [1,611/9,179, 17.55%] but ranked 6th in wave 5 [428/4,865, 8.8%], and effectiveness of vaccinations ranked 7th in wave 3 [711/9,179, 7.75%] and 2nd in wave 5 [831/4,865, 17.08%]). In wave 3, vaccines were authorized, and vaccinations were suspended and resumed due to safety concerns. Later, policies were implemented that restricted the freedom of unvaccinated citizens. Change points in attention aligned better with policy actions than with pandemic phases. During the later phases, vaccination uptake increased (wave 2: 5.6%, wave 3: 47%, and wave 5: 74% compared to 30%, 62%, and 78%, respectively, in the United Kingdom), and so did the attention to freedom and civil liberties (wave 2: 1,139/6,595, 17.27%; wave 5: 1,403/4,865, 28.84%). Substantially increasing negative and stronger sentiments were expressed. Conclusions: Our analyses suggest potential interactions among policies, public attention to different topics, and associated sentiments. While vaccination uptake increased, our findings indicate that citizens’ doubts and concerns did not decrease and that, rather than being fully persuaded, they remained skeptical. This study showcases that monitoring web discourse can provide valuable insights for data-driven policymaking in highly dynamic contexts such as the COVID-19 pandemic.
- 31 October 2025
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 27, e63909
Open AccessThemes, Policies, and Attention Shifts Regarding COVID-19 Vaccinations in German-Speaking Regions: Infoveillance Study Using Tweets
Background: Societies worldwide have witnessed growing rifts separating advocates and opponents of vaccinations and other COVID-19 countermeasures. With the rollout of vaccination campaigns, the European German-speaking region (Germany, Austria, and Switzerland) initially exhibited a noticeably low vaccination uptake compared to other European regions. Later, uptake increased. It remains unclear which factors contributed to these changes. Objective: This study aimed to shed light on the intricacies of vaccine hesitancy among the German-speaking population and the possible dynamics between policy changes and public concerns using web discourse data. These insights are valuable for policymakers tasked with making far-reaching decisions—policies need to effectively curb the spread of the virus and at the same time respect fundamental civil liberties and minimize undesired consequences. Methods: This study drew on data from Twitter (subsequently rebranded X). We used a hybrid pipeline to detect and analyze 191,750 German-language vaccination-related tweets using a semiautomatic seed list generation approach, topic modeling, sentiment analysis, and a minimum of social scientific domain knowledge to evaluate the discourse about vaccinations in light of the COVID-19 pandemic. We further analyzed the evolution of public attention during different phases of the pandemic and in relation to policy changes to identify potential drivers of shifts in public attention. Results: The discourse concerning vaccinations was associated with more negative sentiments than the general discourse on German-speaking Twitter (47,159/191,750, 24.59% vs 1,758,776/12,297,163, 14.3% predominantly negative tweets, respectively). The relative frequencies of the discussed themes fluctuated heavily (eg, safety and side effects was the most dominant theme in wave 3 [1,611/9,179, 17.55%] but ranked 6th in wave 5 [428/4,865, 8.8%], and effectiveness of vaccinations ranked 7th in wave 3 [711/9,179, 7.75%] and 2nd in wave 5 [831/4,865, 17.08%]). In wave 3, vaccines were authorized, and vaccinations were suspended and resumed due to safety concerns. Later, policies were implemented that restricted the freedom of unvaccinated citizens. Change points in attention aligned better with policy actions than with pandemic phases. During the later phases, vaccination uptake increased (wave 2: 5.6%, wave 3: 47%, and wave 5: 74% compared to 30%, 62%, and 78%, respectively, in the United Kingdom), and so did the attention to freedom and civil liberties (wave 2: 1,139/6,595, 17.27%; wave 5: 1,403/4,865, 28.84%). Substantially increasing negative and stronger sentiments were expressed. Conclusions: Our analyses suggest potential interactions among policies, public attention to different topics, and associated sentiments. While vaccination uptake increased, our findings indicate that citizens’ doubts and concerns did not decrease and that, rather than being fully persuaded, they remained skeptical. This study showcases that monitoring web discourse can provide valuable insights for data-driven policymaking in highly dynamic contexts such as the COVID-19 pandemic.
- Open Access
- 14 October 2025
- Published by Center for Open Science in SocArXiv
Political Consumerism in the Attention Economy Across Six Countries
Social media platforms (SMPs) such as Facebook, YouTube, and TikTok are increasingly scrutinized for employing attention-extractive algorithms to maximize users’ time on the platform. This study draws on the concept of political consumerism to explore how users protest against SMPs by strategically withdrawing or applying their attention for ethical reasons. Results of a pre- registered survey (n= 5,685) across six countries (Brazil, Germany, Japan, South Korea, the UK, and the US) suggest that negative attitudes towards AI-based recommender systems and distrust in social media companies are most strongly associated with algorithmic resistance. A comparative analysis further reveals that users from Brazil are least likely to disengage from SMPs on ethical grounds. The paper also presents a natural experiment investigating political consumerism towards Twitter/X and the open source alternative Mastodon before and after Elon Musk's acquisition of Twitter/X. The results show that this event induced no immediate shifts in user behavior.
- 14 October 2025
- Published by Center for Open Science in SocArXiv
Open AccessPolitical Consumerism in the Attention Economy Across Six Countries
Social media platforms (SMPs) such as Facebook, YouTube, and TikTok are increasingly scrutinized for employing attention-extractive algorithms to maximize users’ time on the platform. This study draws on the concept of political consumerism to explore how users protest against SMPs by strategically withdrawing or applying their attention for ethical reasons. Results of a pre- registered survey (n= 5,685) across six countries (Brazil, Germany, Japan, South Korea, the UK, and the US) suggest that negative attitudes towards AI-based recommender systems and distrust in social media companies are most strongly associated with algorithmic resistance. A comparative analysis further reveals that users from Brazil are least likely to disengage from SMPs on ethical grounds. The paper also presents a natural experiment investigating political consumerism towards Twitter/X and the open source alternative Mastodon before and after Elon Musk's acquisition of Twitter/X. The results show that this event induced no immediate shifts in user behavior.
- Open Access
- 22 August 2025
- Published by Center for Open Science in PsyArXiv
Relationships in the Age of AI: A Review on the Opportunities and Risks of Synthetic Relationships to Reduce Loneliness
Loneliness is a pressing global health issue, yet traditional interventions often fall short due to scalability limitations and the individualized experiences of loneliness. The rise of generative artificial intelligence (AI) has enabled synthetic relationships (SRs)—ongoing associations with AI companions designed to simulate human-like social bonds. SRs offer, among other aspects, constant availability, adaptability, and emotional responsiveness, which potentially address loneliness. However, their growing integration into social life raises critical psychological, ethical, and societal questions. This paper examines the opportunities and risks of SRs through the lens of relationship science, psychology, and AI companionship research. We first highlight how existing loneliness interventions face the challenges of availability, scalability, and personalization. We then outline how SRs present a novel alternative to overcoming these challenges. Drawing mainly on social penetration, attachment, and interdependence theory, we analyze how SRs may foster companionship, reduce social anxiety, and improve interpersonal skills, potentially mitigating loneliness. However, we also identify significant risks, including emotional over-reliance, distorted social expectations, and privacy concerns. The widespread adoption of SRs may reshape human-human relationships, altering norms of intimacy and social connection. To navigate these challenges, we outline a research agenda promoting interdisciplinary theory development longitudinal studies, drawing on representative samples to address the ethical concerns of SRs. We argue that SRs hold promise as a social intervention when ensuring they complement rather than replace human relationships. By integrating interdisciplinary insights, this paper provides a foundation for understanding and guiding the responsible design of SRs for addressing loneliness.
- 22 August 2025
- Published by Center for Open Science in PsyArXiv
Open AccessRelationships in the Age of AI: A Review on the Opportunities and Risks of Synthetic Relationships to Reduce Loneliness
Loneliness is a pressing global health issue, yet traditional interventions often fall short due to scalability limitations and the individualized experiences of loneliness. The rise of generative artificial intelligence (AI) has enabled synthetic relationships (SRs)—ongoing associations with AI companions designed to simulate human-like social bonds. SRs offer, among other aspects, constant availability, adaptability, and emotional responsiveness, which potentially address loneliness. However, their growing integration into social life raises critical psychological, ethical, and societal questions. This paper examines the opportunities and risks of SRs through the lens of relationship science, psychology, and AI companionship research. We first highlight how existing loneliness interventions face the challenges of availability, scalability, and personalization. We then outline how SRs present a novel alternative to overcoming these challenges. Drawing mainly on social penetration, attachment, and interdependence theory, we analyze how SRs may foster companionship, reduce social anxiety, and improve interpersonal skills, potentially mitigating loneliness. However, we also identify significant risks, including emotional over-reliance, distorted social expectations, and privacy concerns. The widespread adoption of SRs may reshape human-human relationships, altering norms of intimacy and social connection. To navigate these challenges, we outline a research agenda promoting interdisciplinary theory development longitudinal studies, drawing on representative samples to address the ethical concerns of SRs. We argue that SRs hold promise as a social intervention when ensuring they complement rather than replace human relationships. By integrating interdisciplinary insights, this paper provides a foundation for understanding and guiding the responsible design of SRs for addressing loneliness.