Is there a correlation between AI use and deception in students?

Deceptive behaviour is common among students. Roig et al. (2005) found that students’ deceptive behaviour is correlated with academic misconduct and weaker performance. The availability of LLM based AI chat apps has led to the emergence of AI based academic misconduct, often in the form of students submitting AI work as their own. This kind of academic cheating does not appear to be related to weaker academic performance (Bancoro, 2024). It is currently unknown whether AI use is related to other deceptive behaviours. This study aims to study whether such a relation exists. 100 students will be asked to complete two deception questionnaires: a Serota grid (Serota et al. 2010) and a lying scenarios questionnaire (Blackhurst et al. 2025). They’ll also be asked to complete two AI literacy questionnaires: AILS (Wang et al., 2022) and “EdAIQ” (Elhassan et al., 2025). The main analyses will include correlation between each of the deception measures and each of the AI literacy measures. The outcomes will give greater insight into students’ perception of AI and the support that students may need in incorporating AI into their learning.

Proposed correlational design with 100 students completing self-report questionnaires measuring deception (Serota grid, lying scenarios) and AI literacy (AILS, EdAIQ), with correlation analyses planned.

Background & Context

The emergence of large language model-based AI tools like ChatGPT has created a new frontier in academic integrity. Students are increasingly integrating Gen-AI into their work flows, with AI-based academic misconduct emerging as students submit AI-generated work as their own. Traditional research has linked students' deceptive behavior with academic misconduct and weaker performance, but early evidence suggests AI-based cheating may not follow this pattern.

What remains unknown is whether students who use AI tools are also more prone to other forms of everyday deception. This preregistered study addresses a critical gap: do students with higher AI literacy engage in more frequent lying behaviors, or is there no relationship between technological fluency with AI and general dishonesty? Understanding this connection could inform educational approaches to both AI literacy training and academic integrity interventions.

Research Design & Methodology

This correlational study proposes to recruit 100 university students who will complete a battery of self-report questionnaires measuring two distinct constructs: deception behaviors and AI literacy. The design allows researchers to examine whether individual differences in AI competency relate to patterns of everyday dishonesty.

Deception will be measured using two established instruments:

• The Serota grid, which asks participants to report the number of lies told in a 24-hour period across different receivers and communication modes
• A lying scenarios questionnaire adapted from Warmelink & McLatchie, presenting hypothetical situations that elicit deceptive decision-making

AI literacy will be assessed through two questionnaires:

• AILS (AI Literacy Scale) by Wang et al. (2022), a 12-item instrument measuring awareness, use, evaluation, and ethics
• EdAIQ (Education AI Questionnaire) by Elhassan et al. (2025), focused on AI competencies in educational contexts

The primary statistical approach will involve correlation analyses between each deception measure and each AI literacy measure, providing multiple perspectives on potential relationships between these domains.

Results & Key Findings

This is a preregistered study that has not yet been conducted. No empirical results are currently available. The study is registered to establish hypotheses and analysis plans prior to data collection, which is a best practice in open science to prevent questionable research practices like p-hacking or HARKing (Hypothesizing After Results are Known).

When completed, the analyses will reveal:

• Whether students with higher AI literacy scores report more or fewer lies on the Serota grid
• Whether AI competency correlates with willingness to lie in specific hypothetical scenarios
• Whether different dimensions of AI literacy (awareness, use, evaluation, ethics) show distinct relationships with deception behaviors
• Whether self-reported lying frequency relates to AI literacy differently than scenario-based deception measures

Discussion & Significance

This study represents an important first exploration into whether technological literacy with AI tools relates to broader patterns of dishonesty. The findings could challenge or confirm emerging narratives about AI use in education. If AI literacy correlates positively with deception, it might suggest that technologically proficient students view AI-assisted work as just another form of strategic deception. Alternatively, no correlation would indicate AI use is a distinct behavior, separate from general dishonesty traits.

The research also bridges two important literatures: deception research and educational technology. Research shows general lying frequency measured by the Serota Grid correlates moderately with other lying frequency measures, but scenario-based lying questionnaires often do not strongly predict general lying behavior. Understanding how AI literacy fits into this complex picture of individual differences in deception could inform more nuanced approaches to academic integrity education.

Limitations & Considerations

As a preregistration, the study design itself presents several methodological considerations. The reliance on self-report measures for both deception and AI literacy may introduce social desirability bias—students may underreport both lying behaviors and unauthorized AI use. Research on lying prevalence has found that most people do not lie on a given day, and the majority of lies are told by a few prolific liars, suggesting the distribution may be highly skewed.

The correlational design cannot establish causality. Even if AI literacy and deception correlate, this could reflect third variables (e.g., risk-taking personality, academic pressure) rather than a direct relationship. The student sample may limit generalizability to other populations. Additionally, AI literacy instruments are relatively new and continue to evolve as AI technologies advance rapidly.

Practical Applications

If completed, this research could inform how universities approach AI literacy training. Should AI competency courses include explicit discussions of academic integrity and ethical use? Do students who are comfortable with AI need different interventions than those who are technologically hesitant? The findings could help educators understand whether AI misconduct stems from general dishonesty traits or represents a novel behavioral pattern.

For students, understanding the relationship between AI use and deception could promote self-reflection about their own behaviors. The research may also inform the development of AI detection and prevention strategies that account for individual differences in both technological skill and honesty dispositions.