Determinants and consequences of AI literacy

How to cite

Kosse, Fabian; Falk, Armin. 2026. "Determinants and consequences of AI literacy." AEA RCT Registry. AEARCTR-0018011. https://doi.org/10.1257/rct.18011-1.0

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

Fabian Kosse* , Armin Falk

* Primary investigator

Registry ID
AEARCTR-0018011
DOI
https://doi.org/10.1257/rct.18011-1.0
First registered
Last updated
Status
Completed
Trial dates
2011-09-01 – 2026-03-01
Countries
Germany ()
Registry record
http://www.socialscienceregistry.org/trials/18011

Abstract

Artificial intelligence (AI) is increasingly shaping educational, occupational, and everyday decision-making environments. Understanding individuals’ competencies in interacting with AI systems is therefore crucial for studying social disparities and behavioral responses to technological change.

In this study, we measure AI-related skills using a short survey-based self-assessment battery. The measure captures individuals’ perceived competencies in understanding, evaluating, and effectively using AI tools. It is designed to provide a compact yet informative assessment of AI skills, making it suitable for analyzing socio-economic correlates and for assessing whether exposure to a mentoring relationship can shape the development of AI-related skills.

Intervention

The intervention is a mentoring program grounded in the concept of informal learning, meaning that it embeds learning processes into everyday activities rather than directly targeting academic achievement. Its central aim is to broaden a child’s horizons through regular social interactions with a new attachment figure.

Experimental design

We recruited participants and their parents from the two cities Cologne and Bonn in Germany. In 2011, we invited all families living in those cities with children born between September 2003 and August 2004 to participate in a mentoring program, as well as one third of families with children born between September 2002 and August 2003 (N=14,451). We truthfully informed parents that, due to capacity constraints, participation in the program was not guaranteed. 1626 families indicated a willingness to participate and answered a short questionnaire including questions on income, education and whether both parents lived in the same household. We focus on those children whose parents met at least one of the following three criteria: (i) Equivalence income of the household is lower than 1065 Euro, corresponding to the 30th percentile of the German income distribution. (ii) Neither parent has a school-leaving degree qualifying for university studies. (iii) Parents do not live in the same household. We invited these children (N=700) and their parents for a baseline interview conducted in September to October 2011. 590 children and their parents participated in the baseline interview and gave their written consent to allow the transmission of their address to the organization running the mentoring program. This is our main sample. Out of this sample, 212 families were randomly selected to be treated (“treatment group”), the remaining 378 families form the control group.

We also recruited a high socio-economic status (SES) comparison group from the children whose parents did not meet either of the three criteria listed above (N=150 invited, N=122 participated in the baseline interview and gave written consent).

Design details

Randomization method
Randomization done in office by a computer using STATA.
Randomization unit
Individual
Clusters
590 individuals
Observations
590 individuals
Arms
212 children were randomly selected to be treated (“treatment group”), the remaining 378 children form the control group.
Primary outcomes
Our primary outcomes are AI-related skills, measured using a shortened four-item version of the Meta AI Literacy Scale (MAILS; Carolus et al., 2023), capturing the dimensions of AI literacy and AI self-efficacy. Carolus, A., Koch, M. J., Straka, S., Latoschik, M. E., & Wienrich, C. (2023). MAILS – Meta AI literacy scale: Development and testing of an AI literacy questionnaire based on well-founded competency models and psychological change- and meta-competencies. Computers in Human Behavior: Artificial Humans, 1(2), 100014.

Cited by

No citation data available for this registration yet.

Keywords