Mohammed Alsobay

Group decision-making often suffers from uneven information sharing, hindering decision quality. While large language models (LLMs) have been widely studied as aids for individuals, their potential to support groups of users, potentially as facilitators, is relatively underexplored. We present a pre-registered randomized experiment with 1,475 participants assigned to 281 live groups completing a hidden profile task—selecting an optimal city for a hypothetical sporting event—under one of four facilitation conditions: no facilitation, a one-time message prompting information sharing, a human facilitator, or an LLM (GPT-4o) facilitator. We find that LLM facilitation increased information shared within a discussion by raising the minimum level of engagement with the task among group members, and that these gains came at limited cost in terms of participants’ attitudes towards the task, their group, or their facilitator. Whether by human or AI, there was no significant effect of facilitation on the final decision outcome, suggesting that even substantial but partial increases in information sharing were insufficient to overcome the hidden profile effect studied. To support the design and evaluation of LLM-mediated group decision-making systems, we release our data and our experimental platform, the Group-AI Interaction Laboratory (GRAIL), as an open-source tool.

Do users automatically benefit when generative AI models improve, or does realizing those gains require changing how they interact with the technology? We ran two preregistered experiments with 3,750 participants submitting nearly 37,000 prompts across two versions of OpenAI’s DALL-E to find out. The answer depends critically on task structure. In a bounded task with a clear objective, replicating a specific target image, roughly half of the performance gain came from the model itself, and half came from users naturally adjusting their prompts. In an open-ended creative task, designing logos for hypothetical organizations, nearly all of the improvement was attributable to the model alone. We then asked whether automated prompt rewriting, a feature embedded in many commercial AI products, could substitute for this user adaptation. It could not: rewriting modestly helped in the creative task but substantially hurt performance in the replication task, where precise user control mattered most. For organizations deploying AI in precision-oriented workflows, our results point to an underappreciated priority: investing in how people learn to interact with a model may matter as much as the model itself.

Despite decades of research, the conditions under which punishment promotes cooperation remain unclear. Through an integrative experiment varying 14 design parameters of public goods games across 360 experimental conditions (147,618 decisions from 7100 participants), we reveal substantial heterogeneity in punishment effectiveness: Its impact on welfare ranges from 43% improvement to 44% reduction depending on the game parameters. To characterize these patterns, we developed models that outperformed human forecasters in predicting punishment effectiveness in new experiments. Communication emerges as the most important factor, followed by contribution framing (opt out versus opt in), contribution type (variable versus all-or-nothing), game length, and outcome visibility, though these factors often interact. The results reframe the debate from whether punishment works to when it does, demonstrating how integrative experiments enable discovery of generalizable patterns in social phenomena.

Significance: Scientists and managers alike have been preoccupied with the question of whether and, if so, under what conditions groups of interacting problem solvers outperform autonomous individuals. Here we describe an experiment in which individuals and groups were evaluated on a series of tasks of varying complexity. We find that groups are as fast as the fastest individual and more efficient than the most efficient individual when the task is complex but not when the task is simple. We then precisely quantify synergistic gains and process losses associated with interacting groups, finding that the balance between the two depends on complexity. Our study has the potential to reconcile conflicting findings about group synergy in previous work.