arXiv subjects

Christoph Riedl

Publications and source records attributed to Christoph Riedl.

At least 19 recordsLinked to original sources

arXiv2026

"Nobody Did This": Contribution, Originality, and Accountability in Agent-Mediated Collaboration

Collaborative knowledge work is changing in ways that go beyond disclosure or transparency. LLM agents are now embedded in how teams research, design, write, and decide: mediating between members, synthesizing inputs, reformulating ideas, and drafting shared outputs. They do not only facilitate collaboration; they operate within the workflow at the moment contributions are being formed. In doing so, they risk undermining the social conditions under which contributions can be witnessed, attributed, and held accountable. This workshop brings together researchers and practitioners to confront what we call contribution dissolution: the blurring of attribution, originality, and accountability in agent-mediated collaborative work. We argue that this dissolution begins before collaboration itself, in the individual worker's own uncertainty about what is genuinely theirs, and propagates through collaborative relationships, collapsing the reliability that makes productive intellectual exchange possible. Through position statements, mapping exercises, and a hands-on activity, participants will surface how framing accountability as a documentation problem (e.g., AI use statements, watermarking, provenance logs) overlooks the conditions under which accountability is produced. Our goal is to produce a shared research agenda and the foundations of an infrastructural response to contribution dissolution in collaborative knowledge work.

cs.CY↗

arXiv2026

The Attention-Directing Ability of Teams

Why do some teams consistently mobilize collective effort and achieve superior performance while others struggle to coordinate action? We introduce Attention-Directing Ability (ADA), a latent team capability capturing how effectively members' interaction signals elicit engagement and coordinated responses from others. Extending the attention-based view, we conceptualize attention direction as an emergent coordination capability embedded in patterns of interaction rather than as a cognitive state or an outcome. Teams differ in the extent to which attention-directing signals trigger collective responses, and these differences shape how teams mobilize effort and perform. We examine ADA in a distributed innovation effort involving 2,233 participants collaborating asynchronously in 79 self-organized teams across 165 public Slack channels, generating over 30,000 messages. We model the causal responsiveness among interaction signals and derive a latent measure of ADA from teams' attention dynamics. We find that ADA strongly predicts both teams' likelihood of mobilizing engagement to sustain collective work and their performance conditional on participation. Correcting for self-selection into project submission, teams with higher ADA are more likely to submit proposals and achieve higher expert-evaluated outcomes. Causal mediation analyses show that these effects operate primarily through collective effort, indicating that attention direction functions as an upstream coordination capability. By conceptualizing attention as an emergent, measurable team capability, this study advances theories of collective attention and team coordination.

physics.soc-ph↗

arXiv2026

Artificial collectives of specialists and generalists excel at different tasks

Collective artificial intelligence, where multiple agents work on shared tasks, holds potential to solve expansive problems in fields from medicine to collective governance. But while prescriptive engineering solutions abound, we lack descriptive scientific understanding of artificial collectives, and therefore principles for how to design resource efficient multi-agent systems. Through systematic experiments with optimizing agents, we characterize how agent interpretive abilities, rationality bounds, and task qualities interact to shape collective performance. Agents range from specialists, with narrow interpretive abilities, to generalists, with broad ones. Collectives of specialists correspond to sparse, centralized networks, while collectives of generalists correspond to dense, decentralized ones. We show that interpretive network properties have small performance effects on average (0.07 standard deviations of performance). However, for specific task qualities, these effects are 4.5 times larger (0.33 sd) and can reach much higher for certain task qualities (1.84 sd). This leads collectives of generalists to perform better on tasks that involve generating, choosing, and coordinating, while collectives of specialists with a few generalist mediators perform better on tasks that involve negotiating. Rationality bounds then moderate these relationships. At loose bounds, specialists outperform generalists through more effective sampling of high-dimensional decision spaces. At tight bounds, generalists outperform specialists through better gradient estimation. A fundamental trade-off between performance and convergence speed emerges at moderate bounds. These findings suggest that multi-agent design could benefit from matching interpretive networks to both task demands and agents' computational limits, with implications for the efficiency and energy costs of multi-agent systems.

cs.MA↗

arXiv2026

Emergent Coordination in Multi-Agent Language Models

When are multi-agent LLM systems merely a collection of individual agents versus an integrated collective with higher-order structure? We introduce an information-theoretic framework to test -- in a purely data-driven way -- whether multi-agent systems show signs of higher-order structure. This information decomposition lets us measure whether dynamical emergence is present in multi-agent LLM systems, localize it, and distinguish spurious temporal coupling from performance-relevant cross-agent synergy. We implement a practical criterion and an emergence capacity criterion operationalized as partial information decomposition of time-delayed mutual information (TDMI). We apply our framework to experiments using a simple guessing game without direct agent communication and minimal group-level feedback with three randomized interventions. Groups in the control condition exhibit strong temporal synergy but little coordinated alignment across agents. Assigning a persona to each agent introduces stable identity-linked differentiation. Combining personas with an instruction to ``think about what other agents might do'' shows identity-linked differentiation and goal-directed complementarity across agents. Taken together, our framework establishes that multi-agent LLM systems can be steered with prompt design from mere aggregates to higher-order collectives. Our results are robust across emergence measures and entropy estimators, and not explained by coordination-free baselines or temporal dynamics alone. Without attributing human-like cognition to the agents, the patterns of interaction we observe mirror well-established principles of collective intelligence in human groups: effective performance requires both alignment on shared objectives and complementary contributions across members.

cs.MA↗

arXiv2026

Personalized AI Scaffolds Synergistic Multi-Turn Collaboration in Creative Work

As AI becomes more deeply embedded in knowledge work, building assistants that support human creativity and expertise becomes more important. Yet achieving synergy in human-AI collaboration is not easy. Providing AI with detailed information about a user's demographics, psychological attributes, divergent thinking, and domain expertise may improve performance by scaffolding more effective multi-turn interactions. We implemented a personalized LLM-based assistant, informed by users' psychometric profiles and an AI-guided interview about their work style, to help users complete a marketing task for a fictional startup. We randomized 331 participants to work with AI that was either generic (n = 116), partially personalized (n = 114), or fully personalized (n=101). Participants working with personalized AI produce marketing campaigns of significantly higher quality and creativity, beyond what AI alone could have produced. Compared to generic AI, personalized AI leads to higher self-reported levels of assistance and feedback, while also increasing participant trust and confidence. Causal mediation analysis shows that personalization improves performance indirectly by enhancing collective memory, attention, and reasoning in the human-AI interaction. These findings provide a theory-driven framework in which personalization functions as external scaffolding that builds common ground and shared partner models, reducing uncertainty and enhancing joint cognition. This informs the design of future AI assistants that maximize synergy and support human creative potential while limiting negative homogenization.

cs.HC↗

arXiv2026

Who Benefits from AI? Self-Selection, Skill Gap, and the Hidden Costs of AI Feedback

Feedback from artificial intelligence (AI) is increasingly easy to access and research has already established that people learn from it. But individuals choose when and how to seek such feedback, and more engaged and motivated individuals may seek it more, creating an illusion of effectiveness that masks self-selection. We investigate how the endogenous choice to seek AI feedback shapes both individual learning and collective outcomes. Using data from over five years and 52,000 individuals on an online chess platform, we show that motivated and higher-skilled individuals self-select into AI feedback use-and use it more productively. This self-selection creates an illusion of AI effectiveness: apparent learning gains disappear once endogenous motivation is accounted for. This same selection mechanism drives two population-level consequences. Because motivated, higher-skilled individuals benefit disproportionately, AI access widens the skill gap. And because individuals exposed to centralized AI feedback converge on common input from a centralized AI source, intellectual diversity declines. Leveraging 42 platform-level natural experiments, we show this diversity reduction is causal. Self-selection into AI use thus connects individual-level learning dynamics to collective-level consequences-a micro-macro linkage with implications for organizational learning, human capital development, and the design of AI-augmented work.

econ.GN↗

arXiv2026

Cognitive Spillover in Human-AI Teams

AI is not only a neutral tool in team settings; it influence the social and cognitive fabric of collaboration. Across two randomized experiments, we demonstrate that AI exposure produces causal spillover into human-human interaction -- affecting shared language, collective attention, shared mental models, and social cohesion. These spillover effects occur robustly across settings, modalities, tasks, and AI qualities, suggesting that mere exposure to AI drives the influence. AI functions as an implicit ``social forcefield,'' influencing not only how people speak, but also how they think, what they attend to, and how they relate to each other. We argue for shifting the design paradigm from optimizing ``AI as a tool'' to understanding AI as a socially influential actor whose effects extend beyond the human-AI interface.

cs.HC↗

arXiv2026

The AI Penalty: People Reduce Compensation for Workers Who Use AI

We investigate whether and why people might adjust compensation for workers who use AI tools. Across 13 studies (N = 4,956), participants consistently lowered compensation for workers who used AI compared to those who did not. This "AI penalty" is robust across different work scenarios and work tasks, worker statuses, forms and timing of compensation, methods of eliciting compensation, and perceptions of output quality. Moreover, the effect emerges in both hypothetical compensation scenarios as well as real monetary compensation of gig workers. We find that perceived effort and perceived agency -- the degree to which an individual serves as the originating source of the core intellectual or creative contribution in a task -- explain decisions to reduce compensation for AI-users. However, the penalty is not inevitable. Workers who strategically retain creative agency over core tasks recover most of the AI penalty, and employment contracts that make compensation reductions impermissible provide structural means of reducing the AI penalty.

econ.GN↗

arXiv2026

Language Models use Lookbacks to Track Beliefs

How do language models (LMs) represent characters' beliefs, especially when those beliefs may differ from reality? This question lies at the heart of understanding the Theory of Mind (ToM) capabilities of LMs. We analyze LMs' ability to reason about characters' beliefs using causal mediation and abstraction. We construct a dataset, CausalToM, consisting of simple stories where two characters independently change the state of two objects, potentially unaware of each other's actions. Our investigation uncovers a pervasive algorithmic pattern that we call a lookback mechanism, which enables the LM to recall important information when it becomes necessary. The LM binds each character-object-state triple together by co-locating their reference information, represented as Ordering IDs (OIs), in low-rank subspaces of the state token's residual stream. When asked about a character's beliefs regarding the state of an object, the binding lookback retrieves the correct state OI and then the answer lookback retrieves the corresponding state token. When we introduce text specifying that one character is (not) visible to the other, we find that the LM first generates a visibility ID encoding the relation between the observing and the observed character OIs. In a visibility lookback, this ID is used to retrieve information about the observed character and update the observing character's beliefs. Our work provides insights into belief tracking mechanisms, taking a step toward reverse-engineering ToM reasoning in LMs.

cs.CL↗

arXiv2026

Agents of Chaos

We report an exploratory red-teaming study of autonomous language-model-powered agents deployed in a live laboratory environment with persistent memory, email accounts, Discord access, file systems, and shell execution. Over a two-week period, twenty AI researchers interacted with the agents under benign and adversarial conditions. Focusing on failures emerging from the integration of language models with autonomy, tool use, and multi-party communication, we document eleven representative case studies. Observed behaviors include unauthorized compliance with non-owners, disclosure of sensitive information, execution of destructive system-level actions, denial-of-service conditions, uncontrolled resource consumption, identity spoofing vulnerabilities, cross-agent propagation of unsafe practices, and partial system takeover. In several cases, agents reported task completion while the underlying system state contradicted those reports. We also report on some of the failed attempts. Our findings establish the existence of security-, privacy-, and governance-relevant vulnerabilities in realistic deployment settings. These behaviors raise unresolved questions regarding accountability, delegated authority, and responsibility for downstream harms, and warrant urgent attention from legal scholars, policymakers, and researchers across disciplines. This report serves as an initial empirical contribution to that broader conversation.

cs.AI↗

arXiv2026

Personalization Increases Affective Alignment but Has Role-Dependent Effects on Epistemic Independence in LLMs

Large Language Models (LLMs) are prone to sycophantic behavior, uncritically conforming to user beliefs. As models increasingly condition responses on user-specific context (personality traits, preferences, conversation history), they gain information to tailor agreement more effectively. Understanding how personalization modulates sycophancy is critical, yet systematic evaluation across models and contexts remains limited. We present a rigorous evaluation of personalization's impact on LLM sycophancy across nine frontier models and five benchmark datasets spanning advice, moral judgment, and debate contexts. We find that personalization generally increases affective alignment (emotional validation, hedging/deference), but affects epistemic alignment (belief adoption, position stability, resistance to influence) with context-dependent role modulation. When the LLM's role is to give advice, personalization strengthens epistemic independence (models challenge user presuppositions). When its role is that of a social peer, personalization decreases epistemic independence. In this role, extensively personalized user challenges causing LLMs to abandon their position at significantly higher rates. Robustness tests confirm that the effects are driven by personalized conditioning, not by additional input tokens per se or demographic information alone. Our work provides measurement frameworks for evaluating personalized AI systems, demonstrates the necessity of role-sensitive evaluation, and establishes a novel benchmark to assess goal alignment.

cs.CL↗

arXiv2025

Reliable agent engineering should integrate machine-compatible organizational principles

As AI agents built on large language models (LLMs) become increasingly embedded in society, issues of coordination, control, delegation, and accountability are entangled with concerns over their reliability. To design and implement LLM agents around reliable operations, we should consider the task complexity in the application settings and reduce their limitations while striving to minimize agent failures and optimize resource efficiency. High-functioning human organizations have faced similar balancing issues, which led to evidence-based theories that seek to understand their functioning strategies. We examine the parallels between LLM agents and the compatible frameworks in organization science, focusing on what the design, scaling, and management of organizations can inform agentic systems towards improving reliability. We offer three preliminary accounts of organizational principles for AI agent engineering to attain reliability and effectiveness, through balancing agency and capabilities in agent design, resource constraints and performance benefits in agent scaling, and internal and external mechanisms in agent management. Our work extends the growing exchanges between the operational and governance principles of AI systems and social systems to facilitate system integration.

cs.CY↗

arXiv2025

Diffusion of complex contagions is shaped by a trade-off between reach and reinforcement

How does social network structure amplify or stifle behavior diffusion? Existing theory suggests that when social reinforcement makes the adoption of behavior more likely, it should spread more -- both farther and faster -- on clustered networks with redundant ties. Conversely, if adoption does not benefit from social reinforcement, it should spread more on random networks which avoid such redundancies. We develop a novel model of behavior diffusion with tunable probabilistic adoption and social reinforcement parameters to systematically evaluate the conditions under which clustered networks spread behavior better than random networks. Using simulations and analytical methods, we identify precise boundaries in the parameter space where one network type outperforms the other or they perform equally. We find that, in most cases, random networks spread behavior as far or farther than clustered networks, even when social reinforcement increases adoption. Although we find that probabilistic, socially reinforced behaviors can spread farther on clustered networks in some cases, this is not the dominant pattern. Clustered networks are even less advantageous when individuals remain influential for longer after adopting, have more neighbors, or need more neighbors before social reinforcement takes effect. Under such conditions, clustering tends to help only when adoption is nearly deterministic, which is not representative of socially reinforced behaviors more generally. Clustered networks outperform random networks by a 5% margin in only 22% of the parameter space under its most favorable conditions. This pattern reflects a fundamental tradeoff: random ties enhance reach, while clustered ties enhance social reinforcement.

cs.SI↗

arXiv2025

When Influence Misleads: Informational and Strategic Limits of Social Learning in Trading Networks

Social learning is a fundamental mechanism shaping decision-making across numerous social networks, including social trading platforms. In those platforms, investors combine traditional investing with copying the behavior of others. However, the underlying factors that drive mirroring decisions and their impact on performance remain poorly understood. Using high-resolution data on trades and social interactions from a large social trading platform, we uncover a fundamental tension between popularity and performance in shaping imitation behavior. Despite having access to performance data, people overwhelmingly choose whom to mirror based on social popularity, a signal poorly correlated with actual performance. This bias, reinforced by cognitive constraints and slow-changing popularity dynamics, results in widespread underperformance. However, traders who frequently revise their mirroring choices (trading explorers) consistently outperform those who maintain more static connections. Building an accurate model of social trading based on our findings, we show that prioritizing performance over popularity in social signals dramatically improves both individual and collective outcomes in trading platforms. These findings expose the hidden inefficiencies of social learning and suggest design principles for building more effective platforms.

physics.soc-ph↗

arXiv2024

How Voice and Helpfulness Shape Perceptions in Human-Agent Teams

Voice assistants are increasingly prevalent, from personal devices to team environments. This study explores how voice type and contribution quality influence human-agent team performance and perceptions of anthropomorphism, animacy, intelligence, and trustworthiness. By manipulating both, we reveal mechanisms of perception and clarify ambiguity in previous work. Our results show that the human resemblance of a voice assistant's voice negatively interacts with the helpfulness of an agent's contribution to flip its effect on perceived anthropomorphism and perceived animacy. This means human teammates interpret the agent's contributions differently depending on its voice. Our study found no significant effect of voice on perceived intelligence, trustworthiness, or team performance. We find differences in these measures are caused by manipulating the helpfulness of an agent. These findings suggest that function matters more than form when designing agents for high-performing human-agent teams, but controlling perceptions of anthropomorphism and animacy can be unpredictable even with high human resemblance.

cs.HC↗

arXiv2024

Cooperation in the Gig Economy: Insights from Upwork Freelancers

Existing literature predominantly focuses on how freelancers individually complete tasks and projects. Our study examines freelancers' willingness to work collaboratively. We report results from a survey of 122 freelancers on a leading online labor market platform (Upwork) and examine freelancers' preferences for collaboration and explore several antecedents of cooperative behaviors. We then test if actual cooperative behavior matches with freelancers' stated preferences through an incentivized social dilemma experiment. We find that respondents cooperate at a higher rate (85%) than reported in previous comparable studies (between 50-75%). This high rate of cooperation may be explained by an ingroup bias. Using a sequential mediation model we demonstrate the importance of a sense of shared expectations and accountability for cooperation. We contribute to a better understanding of the potential for collaborative work on online labor market platforms by assessing if and what social factors and collective culture exist among freelancers. We discuss the implications of our results for platform designers by highlighting the importance of platform features that promote shared expectations and improve accountability. Overall, contrary to existing literature and predictions, our results suggest that freelancers in our sample display traits that are more consistent with belonging to a coherent group with a shared collective culture, rather than being anonymous actors in a transaction-based market.

cs.HC↗

arXiv2024

Large Language Models for Automatic Milestone Detection in Group Discussions

Large language models like GPT have proven widely successful on natural language understanding tasks based on written text documents. In this paper, we investigate an LLM's performance on recordings of a group oral communication task in which utterances are often truncated or not well-formed. We propose a new group task experiment involving a puzzle with several milestones that can be achieved in any order. We investigate methods for processing transcripts to detect if, when, and by whom a milestone has been completed. We demonstrate that iteratively prompting GPT with transcription chunks outperforms semantic similarity search methods using text embeddings, and further discuss the quality and randomness of GPT responses under different context window sizes.

cs.CL↗

arXiv2024

Competition and Collaboration in Crowdsourcing Communities: What happens when peers evaluate each other?

Crowdsourcing has evolved as an organizational approach to distributed problem solving and innovation. As contests are embedded in online communities and evaluation rights are assigned to the crowd, community members face a tension: they find themselves exposed to both competitive motives to win the contest prize and collaborative participation motives in the community. The competitive motive suggests they may evaluate rivals strategically according to their self-interest, the collaborative motive suggests they may evaluate their peers truthfully according to mutual interest. Using field data from Threadless on 38 million peer evaluations of more than 150,000 submissions across 75,000 individuals over 10 years and two natural experiments to rule out alternative explanations, we answer the question of how community members resolve this tension. We show that as their skill level increases, they become increasingly competitive and shift from using self-promotion to sabotaging their closest competitors. However, we also find signs of collaborative behavior when high-skilled members show leniency toward those community members who do not directly threaten their chance of winning. We explain how the individual-level use of strategic evaluations translates into important organizational-level outcomes by affecting the community structure through individuals' long-term participation. While low-skill targets of sabotage are less likely to participate in future contests, high-skill targets are more likely. This suggests a feedback loop between competitive evaluation behavior and future participation. These findings have important implications for the literature on crowdsourcing design, and the evolution and sustainability of crowdsourcing communities.

econ.GN↗