The Cost-Benefit of Interdisciplinarity in AI for Mental Health

AI summary

Overview

Research area: AI safety and ethics, specifically interdisciplinary collaboration in AI mental health chatbots.

Technical level: Beginner-Friendly. This is a conceptual position paper with no experimental component, no model training, and no quantitative benchmarks.

Scope: The paper argues that technology, healthcare, ethics, and law experts should be deliberately embedded at specific phases of the mental health chatbot lifecycle to achieve value-alignment and comply with EU AI Act requirements for high-risk AI systems.

What This Paper Is About

Most AI mental health chatbots are built from a narrow set of disciplinary perspectives and do not integrate expertise across their entire lifecycle. The authors examine the cost-benefit trade-off of interdisciplinary collaboration, reviewing recent chatbot projects to show how rarely such collaboration actually occurs and how limited it is when it does. Their goal is to argue for a phased model of interdisciplinary input and to offer practical recommendations for making that model work.

Key Contributions

1. A review of interdisciplinary collaboration in recent mental health chatbots, drawing on a 2023 survey finding that only 19.6% of chatbots employed interdisciplinary teams, plus the authors' own overview table of seven chatbots.

2. A position statement that technology, healthcare, ethics, and law experts should be deliberately embedded in the most impactful lifecycle phases (design, development, evaluation), grounded in the EU AI Act's classification of mental health AI chatbots as high-risk.

3. A phased collaboration model (Figure 1) describing what each expert group contributes: technology experts guide implementation with HCI principles, human-in-the-loop methods, and user-centric evaluation; mental healthcare professionals identify user needs and evaluate benefits, risks, and user satisfaction; ethicists define ethical and social values and ensure value-alignment; legal advisors specify regulatory requirements and ensure compliance.

4. Practical recommendations covering interdisciplinary methodologies and frameworks, AI Act compliance, and expert values and reflection.

Main Findings

• Interdisciplinarity is rare in practice: A 2023 survey on conversational agents in mental health found that only 19.6% of chatbots employed interdisciplinary teams.

• The authors' own overview confirms narrow collaboration: Using convenience sampling on papers from 2022-2025 from MDPI (with "chatbot" and "mental health" in the title or keywords) plus other papers explicitly applying interdisciplinary collaboration, the authors screened 16 abstracts. Their overview table presents seven chatbots.

• Some chatbots use only one discipline: Manole et al. (2024) included healthcare experts but not technology, ethics, or law; Kamdan et al. (2025) included technology experts but not healthcare, ethics, or law.

• Most reviewed chatbots pair only technology and healthcare: Hall et al. (2022), Noble et al. (2022), and Chua et al. (2023) included technology and healthcare experts across all lifecycle phases, with no ethics or law input.

• Ethics input, when present, is often phase-limited: Rathnayaka et al. (2022) integrated technology, healthcare, and ethics experts, but only in the design phase.

• Full-lifecycle, multi-discipline collaboration is possible but exceptional: Olla et al. (2025) extended technology, healthcare, and ethics collaboration throughout the lifecycle, and is the only example in the table with ethics involvement in all phases. No reviewed chatbot included law experts.

• The trade-off is real: Even where interdisciplinary teams exist, collaboration tends to be limited to specific phases and does not cover a wide range of disciplines.

• Barriers are documented: Challenges include misaligned expectations, limited cross-disciplinary training, conflicting timelines, tensions between business and scientific standards, communication barriers such as different terminology and professional priorities, restricted funding across fields, and publishing norms that discourage interdisciplinary work.

• The regulatory case is explicit: The EU AI Act classifies mental health AI chatbots as high-risk and sets strict requirements for transparency, human oversight, and risk management, which the authors argue requires technology, healthcare, and law expertise to translate into practice.

• Ethical frameworks support the argument: Value Sensitive Design and ethics-by-design are presented as practical methods for integrating ethical and social values across the lifecycle and anticipating risks in high-risk AI systems.

Methodology in Plain English

This is a position paper, so the work is argumentative rather than experimental. The authors first review the literature on mental health chatbots and the disciplines needed to build them. They then build a small overview table of recent chatbots using convenience sampling: papers from 2022-2025 from MDPI with "chatbot" and "mental health" in the title or keywords, chosen to build on the 2023 survey by Cho et al., plus other papers from different publishers in which chatbots explicitly applied interdisciplinary collaboration. After abstract screening of 16 papers, seven chatbots were included in the table.

For each chatbot, the table records which expert groups (technology, healthcare, ethics, law) contributed, marked with the first letter of each group, and in which lifecycle phase (design, development, evaluation). The authors then interpret this table, describe the barriers to interdisciplinarity from the literature, state their position, and sketch a phased collaboration model in Figure 1. No data collection from users, no chatbot deployment, and no quantitative evaluation were performed.

Why This Matters

Impact on research: The paper reframes interdisciplinarity from a vague aspiration into a design decision with costs and benefits that should be analyzed per lifecycle phase. It connects AI ethics literature to concrete regulatory obligations under the EU AI Act, and it points out that the field currently lacks empirical analysis of the trade-off.

Real-world applications:

• Guiding teams building mental health chatbots on which experts to involve at design, development, and evaluation stages.
• Helping developers of high-risk AI systems document transparency, human oversight, and risk management to meet EU AI Act requirements.
• Informing institutional and funding decisions about supporting cross-sectoral research, given the structural barriers the paper identifies.
• Providing health systems and clinical stakeholders with a framework for evaluating whether a chatbot's development process included adequate clinical and ethical input.

Industry relevance: Companies deploying or procuring mental health chatbots face regulatory obligations as high-risk AI providers. The paper suggests that legal and ethical expertise cannot be treated as a late-stage checkbox but must be embedded in the phases where it has the most impact, which has direct implications for team composition, project timelines, and development cost.

Future Directions

• Empirical analysis of the trade-off: The authors explicitly identify empirical analysis of the cost-benefit trade-off of involving different experts across the chatbot lifecycle as a critical future research dimension, necessary to develop concrete practical recommendations.

• Tangible interdisciplinarity measurement: The paper recommends a measurable interdisciplinarity evaluation measure such as Rao-Stirling, and cross-sectoral methodologies, but does not apply one.

• Adaptation to specific conditions: The position addresses the broad mental health domain rather than specific disorders. The authors note that each condition may require additional expertise, giving the example of involving psychiatry professionals with schizophrenia expertise for chatbots targeting adults with schizophrenia.

• Expanding the expert set: The authors acknowledge the field is evolving and that additional experts may be needed beyond the four groups they propose.

Target Audience

Researchers and practitioners working at the intersection of AI, mental health, and ethics; developers and product teams building mental health chatbots; clinical and psychiatric professionals involved in digital mental health; ethicists and legal experts advising on AI Act compliance; and policy or funding bodies interested in how interdisciplinary collaboration in high-risk AI is structured and supported.

Authors’ abstract

Artificial intelligence has been introduced as a way to improve access to mental health support. However, most AI mental health chatbots rely on a limited range of disciplinary input, and fail to integrate expertise across the chatbot's lifecycle. This paper examines the cost-benefit trade-off of interdisciplinary collaboration in AI mental health chatbots. We argue that involving experts from technology, healthcare, ethics, and law across key lifecycle phases is essential to ensure value-alignment and compliance with the high-risk requirements of the AI Act. We also highlight practical recommendations and existing frameworks to help balance the challenges and benefits of interdisciplinarity in mental health chatbots.