Build games

BUGS investigates advice-taking at different levels of goal-directed action. Before each block, players can choose a human-like advisor, a robot-like advisor, or no advisor. Each trial involves selecting a valley, a cloud of bugs, and the keys used to move. Advice can target distal goals, proximal goals, or motor actions, with experimentally varied reliability and optional explanations. Choices can be made freely or externally imposed. The task measures whether players follow the advice, their reaction times, their ratings of the advisor's acceptability and human-likeness, and their own sense of agency. It tests whether AI advice is accepted less readily, whether higher-level advice carries more weight, and whether unreliable AI advisors are judged more harshly.

• Le Bars, Devaux, Nevidal, Chambon & Pacherie (2020). Agents' pivotality and reward fairness modulate sense of agency in cooperative joint action. Cognition.
• Dubourg, Chambon & Baumard (2025). Human motivation is organized hierarchically, from proximal (means) to ultimate (ends). Behavioral and Brain Sciences.

Meta-choice means choosing whether to choose. The experiment examines whether people value control even when delegation could yield the same reward, and whether this preference for choice varies with risk attitudes and environmental controllability. In a 3D ball game, participants interact with virtual players over a configurable sequence of throws. They can choose who receives the ball or delegate that decision. Researchers configure the reward conditions, publish the experiment, and share a session URL through their recruitment platform. Choices, outcomes, and timestamps are exported for analysis. The design brings together research on the intrinsic value of choice, controllability, and the sense of agency in cooperative action.

• Munuera et al. (2023). Intrinsic motivation for choice varies with individual risk attitudes and the controllability of the environment. PLOS Computational Biology.
• Le Bars, Devaux, Nevidal, Chambon & Pacherie (2020). Agents' pivotality and reward fairness modulate sense of agency in cooperative joint action. Cognition.
• Katiyar, T. (2024). Social Ostracism and Agency in the Digital World, Master's thesis, ENS-PSL.

DEEP conceptualizes player agency as a hierarchy of goals that differ in their level of abstraction and the exploration they require. This framework identifies four key dimensions of player agency: Discovering, Experimenting, Expanding, and Performing. An initial study used factor analysis to examine ratings of game-related actions from 750 participants. A second study tested the proposed hierarchy using large language model annotations of 16,000 video games. A preregistered study with 1,000 participants found that questionnaire scores predict game preferences and are associated with personality and age. The web app administers the questionnaire and provides a score for each dimension. This work now informs the development of new games exploring agency and choice within the ANR-funded MAGIC project.

• Dubourg & Chambon (2025). DEEP: A model of gaming preferences informed by the hierarchical nature of goal-oriented cognition. Entertainment Computing.
• Dubourg, Chambon & Baumard (2025). Human motivation is organized hierarchically, from proximal (means) to ultimate (ends). Behavioral and Brain Sciences.

FORMIDABLE tests whether fear responses are driven by distinct inputs rather than a single threat signal. Eight factors can be varied independently: monster formidability, threat imminence, predictability, predator resemblance, anomaly, hybridity, disgust, and intentionality. About 40 continuous morphological parameters map these factors onto procedurally generated monsters, making the space of possible monsters part of the experimental design. Monster features are assigned independently of the player. Fear is inferred from behaviors that incur time or effort costs without yielding rewards, including response latency under imminent threat, protective actions, glances, and avoidance. Ratings embedded in the game's narrative complement these measures, and participants complete a Big Five personality questionnaire at the start.

• Dubourg & Scrivner (2026). Vulnerability and the computational logic of fear: insights from the horror genre. Evolution and Human Behavior.
• Scrivner & Clasen (2022). Why frightening imaginary worlds? Morbid curiosity and the learning potential of horror. Behavioral and Brain Sciences.