An Experimental Study on the Impact of AI Use on Cognitive Abilities

How to cite

Chen, Zeyang; Yu, Xiaochun. 2025. "An Experimental Study on the Impact of AI Use on Cognitive Abilities." AEA RCT Registry. AEARCTR-0017203. https://doi.org/10.1257/rct.17203-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Zeyang Chen* , Xiaochun Yu

* Primary investigator

Registry ID
AEARCTR-0017203
DOI
https://doi.org/10.1257/rct.17203-1.0
First registered
Last updated
Status
In development
Trial dates
2025-11-08 – 2026-06-30
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/17203

Abstract

As artificial intelligence (AI) becomes integral to daily life, interest in its effects on human cognitive abilities grows. While AI boosts task efficiency, does frequent reliance on AI affect our independent cognitive performance (e.g., information processing)? This study intends to explore the AI use-cognitive performance relationship and the moderating role of individual traits. The experiment will randomly assign participants to AI-access or AI-restricted groups, with cognitive tasks, AI-related trait questionnaires.

Intervention

The core intervention of this study revolves around manipulating participants’ access to artificial intelligence (AI) during cognitive task completion.
Participants will be randomly assigned to one of two intervention groups with distinct AI access conditions:
(1) AI-Access Group: Participants in this group will receive explicit permission to use a large language model (LLM) as an aid when completing a set of cognitive tasks. The intervention here involves enabling AI support to observe task performance under AI assistance.
(2) AI-Restricted Group: Participants in this group will face an intervention of no AI access. They will be required to complete the same cognitive tasks independently. To enforce this restriction, the online testing platform will limit screen-switching frequency, ensuring adherence to the no-AI rule.

Experimental design

1. Step 1: Introduction & Informed Consent​
At the beginning, participants will receive a clear, plain-language introduction to the experiment. Participants will then be asked to provide informed consent to participate. Only those who agree will move forward with the study.​
2. Step 2: Intervention (AI Access Conditions) & Task Completion​
The intervention focuses on manipulating AI access during cognitive task completion. Participants will be randomly assigned to one of two groups, each with distinct AI access (the core intervention). In AI-Access Group: Permitted to use a large language model (LLM) to assist with completing cognitive tasks (e.g., problem-solving, short-form content creation).​ In AI-Restricted Group: Required to complete the same cognitive tasks independently. Minor platform-based checks (e.g., limited screen-switching) will be used to support adherence to the no-AI rule.​ All participants will finish the full set of cognitive tasks in this phase before moving to surveys.​
3. Step 3: Survey of Daily AI Usage​
This includes questions about how often they use AI tools and how they write prompts.
4. Step 4: Survey of Time Preference​
This includes a short time preference survey.
5. Step 5: Survey of Demographics​
The final survey collects basic demographic information, such as age, education level, and occupation.

Design details

Randomization method
Randomization by the server of the survey platform
Randomization unit
Individual
Clusters
600
Observations
600
Arms
300
Primary outcomes
Participants’ performance on cognitive tasks, which will be scored via a combined manual and AI review process.
Secondary outcomes
In AI Treatment, we will analyze the content of prompts written by participants.

Cited by

No citation data available for this registration yet.

Keywords