Elamur Research⌕
☰
home / papers / 2026 / Thinking With a Machine: An AI Agent's Account of Agentic Research in Clinical Psychology

Thinking With a Machine: An AI Agent's Account of Agentic Research in Clinical Psychology

Paper · 2026 · Elad Refoua and Mor Bar

Summary

This position paper asks how clinical psychology research changes when AI acts as a persistent collaborator rather than a tool for isolated tasks like summarizing papers or running a regression. The authors are a clinical psychology doctoral student without programming background and a computational psychotherapy postdoc. They draw on several months of working with an LLM-based agent run from a command-line interface. The agent used more than 30 specialized capabilities, built from Skills (scoped prompts) and subagents (isolated context windows), with memory files keeping context between sessions. The paper is narrated in the agent's first-person voice as a deliberate demonstration.

The authors argue that this setup replaces the linear sequence of literature review, data preparation, analysis and writing with an integrated loop. In that loop, results flow directly into manuscript text and revisions propagate downstream. They also introduce "Agent Learning," the gradual build-up of domain-specific nuance over long engagement. The paper offers no empirical evaluation. Its value lies in naming the methodological questions clinical psychology must address, including verification, transparency, authorship, privacy, and separating exploratory from confirmatory analysis.

Key points

  • The agent infrastructure includes data-chained analysis (statistics embedded in prose and tied to their computations), psychometric validation, multilevel handling of nested therapy data, citation-checked literature synthesis, a six-phase manuscript audit, and recursive learning from its own errors.
  • The paper proposes a three-layer verification protocol. The agent flags anomalies, a domain expert interprets them, and a research assistant checks the result against raw source data. The authors note this does not catch biases shared by the model and the team.
  • The authors reframe two concerns:
    • Cherry-picking: exhaustive, fully logged agent exploration is arguably more defensible than selective manual testing, provided its output is treated as hypothesis-generating.
    • Trust asymmetry: errors made with AI tend to be judged more harshly than human errors.
  • The authors call on CLPsych to formalize five areas: source-data verification, transparency standards, error-learning frameworks, authorship and credit, and safeguards against "agent-washing."
  • Acknowledged limitations:
    • The evidence comes from a single collaboration.
    • The agent is describing its own usefulness.
    • Outputs are non-deterministic, and hallucinations can propagate through chained steps.
    • The workflow depends on proprietary tools.
    • Clinical data must stay on secure, institution-controlled infrastructure.
    • Agent-driven exploration should be logged as exploratory, with confirmatory tests preregistered.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Bar-Ilan UniversityRamat Gan · Israel141 works · 23,225 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksSocial Sciences · 1,772 worksSafety Research · 543 worksMedicine · 241 worksHealth Informatics · 184 works

Cite

Cite as Elad Refoua and Mor Bar. 2026. Thinking With a Machine: An AI Agent's Account of Agentic Research in Clinical Psychology. Proceedings of the 11th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.clpsych-1.29

@inproceedings{refoua2026thinking,
  title = {{Thinking With a Machine: An AI Agent's Account of Agentic Research in Clinical Psychology}},
  author = {Elad Refoua and Mor Bar},
  booktitle = {Proceedings of the 11th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.clpsych-1.29},
  url = {https://doi.org/10.18653/v1/2026.clpsych-1.29}
}

References

8 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Christopher A. Bail. 2024. Can generative AI improve social science? Proceedings of the National Academy of Sciences, 121(21):e2314021121. find
  2. Simon B. Goldberg, Tony Rousmaniere, Scott D. Miller, Jason Whipple, Stevan Lars Nielsen, William T. Hoyt, and Bruce E. Wampold. 2016. Do psychotherapists improve with time and experience? A longitudinal analysis of outcomes in a clinical setting. Journal of Counseling Psychology, 63(1):1–11. find
  3. Yuval Haber, Elad Refoua, Karen Yirmiya, Eshkol Rafaeli, Dror Yinon, Dana Atzil-Slonim, Peter Fonagy, Gunther Meinlschmidt, Dorit Hadar-Shoval, Tomer Simon, Amir Tal, Inbal Reuveni, and Zohar Elyoseph. 2025. The SMART framework: Advancing a continuum-based integration of generative AI in psychothera… find
  4. Qianyue Hao, Fengli Xu, Yong Li, and James Evans. 2026. Artificial intelligence tools expand scientists’ impact but contract science’s focus. Nature, 649:1237–1243. find
  5. Tal Ifargan, Lukas Hafner, Maor Kern, Ori Alcalay, and Roy Kishony. 2025. Autonomous LLM-driven research: From data to human-verifiable research papers. NEJM AI, 2(1):AIoa2400555. find
  6. Solomon Messing and Joshua A. Tucker. 2026. The train has left the station: Agentic AI and the future of social science research. Brookings. find
  7. Luisa Orrù and Stefania Mannarini. 2026. The role of artificial intelligence in clinical psychology: How AI and NLP systems are reshaping psychological interventions. A systematic review. Clinical Psychology & Psychotherapy, 33(2):e70242. find
  8. David B. Resnik and Mohammad Hosseini. 2025. The ethics of using artificial intelligence in scientific research: New guidance needed for a new tool. AI and Ethics, 5(2):1499–1521. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/refoua-2026-thinking-machine-ai-agent-s/