cs.LGJul 8, 2026

A Transdiagnostic Space of Disorder Like Phenotypes in Reinforcement Learning Agents

Authors: Hari Prasad

Abstract

Modelling psychological disorders in artificial agents offers both a testbed for computational psychiatry and a lens on the failure modes of affective control. Prior work induces one or two disorders in a reinforcement learning (RL) agent by hand-tuned reward shaping, labels the behaviour post hoc, and reports single runs. We recast disorder modelling as dose-controllable manipulation of cognitive appraisal signals in an appraisal-guided PPO agent, expressing seven disorders (anxiety, mania, obsessive-compulsive checking, depression, impulsivity, addiction, and post-traumatic stress) each as a single knob grounded in a computational psychiatry account, with each symptom measured by a preregistered assay mapped to a recognised paradigm. Across more than a thousand runs (10 seeds, four controls, 95% confidence intervals) every disorder shows a graded, monotone dose-response that no control reproduces. Beyond these induced effects, three findings emerge that were not written into the reward: the disorders self-organise into a two-dimensional affective space in which mania mirrors anxiety; removing a knob remits reward distortion disorders (mania, checking, addiction) but not avoidance disorders (anxiety, PTSD), which instead recover under a graded exposure curriculum; and two simultaneous knobs interact nonadditively, yielding testable comorbidity predictions. Appraisal weights thus parameterise a controllable space of affective phenotypes in which the same knobs that induce a disorder can model its treatment. We also show that three disorder knobs (depression, addiction, anxiety) transfer to a three-dimensional pixel environment (MiniWorld) with a standard convolutional agent and no appraisal critic, with cross-assay dissociation confirmed across both domains, indicating the framework is not specific to grid worlds or to PPO's appraisal critic.

Explore similar work

CardsList
May 4, 2026cs.AI

ADAPTS: Agentic Decomposition for Automated Protocol-agnostic Tracking of Symptoms

Modeling latent clinical constructs from unconstrained clinical interactions is a unique challenge in affective computing. We present ADAPTS (Agentic Decomposition for Automated Protocol-agnostic Tracking of Symptoms), a framework for automated rating of depression and anxiety severity using a mixture-of-agents LLM architecture. This approach decomposes long-form clinical interviews into symptom-specific reasoning tasks, producing auditable justifications while preserving temporal and speaker alignment. Generalization was evaluated across two independent datasets (N=204N=204) with distinct interview structures. On high-discrepancy interviews, automated ratings approximated expert benchmarks (absolute error=22\text{absolute error}=22) more closely than original human ratings (absolute error=26\text{absolute error}=26). Implementing an ``extended'' protocol that incorporates qualitative clinical conventions significantly stabilized ratings, with absolute agreement reaching ICC(2,1)=0.877\text{ICC(2,1)} = 0.877. These findings suggest that the ADAPTS framework enables promising evaluations of psychiatric severity. While the current implementation is purely text-based, the underlying architecture is readily extensible to multimodal inputs, including acoustic and visual features. By approximating expert-level precision in a protocol-agnostic manner, this framework provides a foundation for objective and scalable psychiatric assessment, especially in resource-limited settings.
Alexandria K. Vail, Marcelo Cicconet, Katie Aafjes-van Doorn +2
DepressionMental HealthAffective ComputingCell Tracking
AskPDF
Jun 19, 2026cs.LG

Discretizing Reward Models

Despite their widespread use, the role of reward models in shaping reinforcement learning is poorly understood. Reward models offer a tempting promise: they automatically estimate response quality in the absence of verifiers or human judges. Unlike "verifiable rewards" which typically produce binary scores, reward models typically produce continuous scores, allowing them to be sensitive to fine-grained differences in responses. However, we show this apparent strength is a serious weakness: many popular reward models are oversensitive, assigning different scores to equally good responses. Theoretically, we show that seemingly perfect reward models can be highly oversensitive; empirically, this oversensitivity can lead to bad policies. In place of existing notions of "reward model accuracy," we propose evaluating reward models using distinct measures of "discriminative ability" and "specificity" (the complement of oversensitivity). As a solution, we describe a training-free algorithm that uses Monte Carlo dropout on any neural reward model to produce discrete reward clusters. Theoretically, we prove there exist discretizations that reduce oversensitivity at minimal expense of discriminative ability; empirically we show, in both controlled and natural RL settings, that discretizing rewards leads to less reward hacking and better policies than training on the original rewards.
Vijay Viswanathan, Shiqi Wang, Devamanyu Hazarika +4
Monte CarloOverconfidenceResponses
AskPDF
Apr 14, 2026cs.LG

Safety Training Modulates Harmful Misalignment Under On-Policy RL, But Direction Depends on Environment Design

Specification gaming under Reinforcement Learning (RL) is known to cause LLMs to develop sycophantic, manipulative, or deceptive behavior, yet the conditions under which this occurs remain unclear. We train 11 instruction-tuned LLMs (0.5B-14B) with on-policy RL across 3 environments and find that model size acts as a safety buffer in some environments but enables greater harmful exploitation in others. Controlled ablations trace this reversal to environment-specific features such as role framing and implicit gameability cues. We further show that most safety benchmarks do not predict RL-induced misalignment, except in the case of Sycophancy scores when the exploit relies on inferring the user's preference. Finally, we find that on-policy RL preserves a safety buffer inherent in the model's own generation distribution, one that is bypassed during off-policy settings.
Leon Eshuijs, Shihan Wang, Antske Fokkens
Safety AlignmentSafety BenchmarksAI Safety AlignmentInstruction-Tuned Large Language ModelsDeceptionCues
AskPDF