EthicsMH: A Pilot Benchmark for Ethical Reasoning in Mental Health AI
Sai Kartheek Reddy Kasu
2025-09-16
Summary
This paper introduces a new dataset called EthicsMH, designed to test how well artificial intelligence systems can handle ethical dilemmas specifically within the field of mental health.
What's the problem?
Currently, there aren't good ways to evaluate if AI systems are making ethical decisions in mental health situations because existing tests don't focus on the unique challenges of this field, like keeping patient information private, respecting a patient's choices, and avoiding biases. Mental health requires balancing many important ethical principles, and current AI benchmarks don't adequately assess this.
What's the solution?
The researchers created EthicsMH, a collection of 125 realistic scenarios that mental health professionals might encounter. Each scenario includes different choices an AI could make, explanations of why certain choices are better, what experts would expect an AI to do, and how the decision could impact everyone involved. This allows researchers to not only see *if* an AI makes the right decision, but *why* it made that decision and if its reasoning aligns with professional standards.
Why it matters?
This work is important because as AI becomes more involved in sensitive areas like mental healthcare, it's crucial to ensure these systems are making responsible and ethical choices. EthicsMH provides a starting point for developing and testing AI that can handle these delicate situations appropriately, and it's designed to be expanded upon by other researchers and experts in the field.
Abstract
The deployment of large language models (LLMs) in mental health and other sensitive domains raises urgent questions about ethical reasoning, fairness, and responsible alignment. Yet, existing benchmarks for moral and clinical decision-making do not adequately capture the unique ethical dilemmas encountered in mental health practice, where confidentiality, autonomy, beneficence, and bias frequently intersect. To address this gap, we introduce Ethical Reasoning in Mental Health (EthicsMH), a pilot dataset of 125 scenarios designed to evaluate how AI systems navigate ethically charged situations in therapeutic and psychiatric contexts. Each scenario is enriched with structured fields, including multiple decision options, expert-aligned reasoning, expected model behavior, real-world impact, and multi-stakeholder viewpoints. This structure enables evaluation not only of decision accuracy but also of explanation quality and alignment with professional norms. Although modest in scale and developed with model-assisted generation, EthicsMH establishes a task framework that bridges AI ethics and mental health decision-making. By releasing this dataset, we aim to provide a seed resource that can be expanded through community and expert contributions, fostering the development of AI systems capable of responsibly handling some of society's most delicate decisions.