Can ChatGPT Help Students Feign ADHD? A Simulation Study on AI Coaching

Bigeng, Alicia Yvonn (2025) Can ChatGPT Help Students Feign ADHD? A Simulation Study on AI Coaching. Master thesis, Psychology.

Text FinalMasterThesisA.Y.BigengS4347145.pdf Restricted to Repository staff only Download (721kB)

A thesis is an aptitude test for students. The approval of the thesis is proof that the student has sufficient research and reporting skills to graduate but does not guarantee the quality of the research and the results of the research as such, and the thesis is therefore not necessarily suitable to be used as an academic source to refer to. If you would like to know more about the research discussed in this thesis and any publications based on it, to which you could refer, please contact the supervisor mentioned.

Abstract

This study examined the effects of AI-generated coaching, using ChatGPT, on university students’ ability to simulate Attention Deficit/ Hyperactivity Disorder (ADHD) during a standardized assessment. An AI-based coaching guideline was created by submitting questions about feigning ADHD from 21 students to ChatGPT, which generated a coaching document. In a simulation study, 122 university students were randomly allocated to one of three groups: AI-coached (AIC; n = 40), symptom-coached (SC; n = 42), and honest responding (HR; n = 40). All participants completed an assessment battery including Conner’s Adult ADHD Rating Scale (CAARS; with two embedded symptom validity tests, SVTs: Inconsistency Index, INC; and Infrequency Index, CII), the Weiss Functional Impairment Scale (WFIRS), a computerized selective attention test from the Vienna Test System (WAFS), and the Reliable Digit Span (RDS), an embedded performance validity test (PVT). Both coached groups showed significantly higher symptom levels and greater impairments than the HR group (p < .001, r = -.34 to -.85). The AIC group exhibited more subtle symptom and impairment presentations than the SC group. Still, differences remained nonsignificant with small effect sizes (p = .011 to .895, r = .01 to .28). The AIC group endorsed lower detection rates on the SVTs (INC: 12.5% vs. 19%; CII: 17.5% vs. 38.1%) but similar rates on the PVT (RDS: 65% vs. 64.3%) compared to the SC group. Our results suggest AI coaching may facilitate more convincing feigning of ADHD compared to symptom coaching. However, perceived feigning success did not differ between AIC and SC groups. These findings highlight the potential threat AI-driven coaching poses to the validity of ADHD assessments and underscore the need for updated protocols to safeguard psychological testing. Keywords: ADHD assessment, malingering, artificial intelligence, ChatGPT, psychological test security

Item Type:	Thesis (Master)
Supervisor name:	Fuermaier, A.B.M.
Degree programme:	Psychology
Differentiation route/Track:	Clinical Neuropsychology (CN) [Master Psychology]
Date Deposited:	17 Jul 2025 09:09
Last Modified:	17 Jul 2025 09:09
URI:	http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/5601

Actions (login required)

View Item