AI evidence extraction

Balancing regulation and innovation: the need for agile AI governance in higher education – a cross-country study

Eyüp Şen, David Vaněček, Müge Adnan · Studies in Higher Education · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1080/03075079.2026.2614986

Methodology & findings

Study design

Mixed-method case study combining convergent parallel design with concurrent qualitative and quantitative data collection.

Sample

N = 104, 2 groups

Primary method

Descriptive statistics (means and standard deviations) to compare responses across institutions. Internal reliability assessed using Cronbach's alpha (α = .902 for all nine items; α = .866 for Adaptation and Flexibility; α = .892 for Progressive Receptiveness; α = .728 for Transparency and Collaboration). Mann-Whitney U tests conducted to verify whether descriptive differences between institutions were statistically significant. Thematic analysis used for open-ended responses with iterative coding and refinement. Convergent parallel mixed-methods design with separate analysis and then merged interpretation. Statistical software and analysis platform not explicitly stated.

Main result

The study found that "MSKU participants reported relatively higher scores on dimensions such as Flexibility, Autonomy, and Adaptability, indicating a perceived openness to change despite institutional concerns raised in their open-ended responses," while "CTU participants expressed more balanced but consistently moderate evaluations across all dimensions, suggesting a more stable yet cautious perception of institutional agility." The research reveals that "both binding and non-binding regulations can constrain effective AI governance. In CTU's case, a structured regulatory framework appears to limit institutional adaptability and iterative learning, leading to modest scores in areas such as flexibility and continuous improvement. However, MSKU's flexibility is undermined by a lack of clear guidelines, user-centric designs and participatory mechanisms vital for agile governance."

Reports effect sizes.

Research paradigm

Mixed methods (qualitative and quantitative); pragmatist/interpretivist approach to policy analysis

Author conclusions

The authors conclude: "This research offers a comparative exploration of how HEIs in different regulatory environments are responding to the complex challenges posed by AI... The results highlight that both binding and non-binding regulations can constrain effective AI governance." They propose that "Agile governance provides institutions with a pragmatic framework to respond proactively to the rapid evolution of AI, while ensuring ethical responsibility and institutional legitimacy." Furthermore, "HEIs must avoid reactive management models that either delay necessary adoption or rush unprepared into technological dependency. Instead, what is needed is a deliberate and reflective approach, one that centers on agility, institutional values, and long-term readiness. By embedding agile thinking into policy design, HEIs can better navigate the tension between innovation and control, empower academic communities, and prepare students for an AI-augmented future."

Risk of bias

Selection bias: Participants self-selected based on prior interest in AI governance (required confirmation of sufficient AI knowledge/experience); Response bias: Reliance on self-reported perceptions of policy agility may not reflect actual implementation; Sampling bias: Purposive sampling of only two institutions limits generalizability; Unmeasured confounding: Cultural dimensions such as academic freedom were not explicitly considered; Selection bias from purposive sampling of participants with prior interest in AI governance; Self-report bias in perception of policy agility and effectiveness; Small sample size (104 academics from two institutions only); Participant self-selection based on self-reported familiarity with AI regulations; Unmeasured cultural dimensions such as academic freedom that may influence responses; Selection bias: Purposive sampling limited to self-reported familiarity with AI governance, creating prior interest bias; Participant bias: Only 104 participants across two institutions, limiting generalizability; Measurement bias: Self-reported perceptions of policy agility may not reflect actual implementation practices; Institutional context bias: Participants only from two specific universities in Czechia and Türkiye; Unmeasured confounding: Cultural dimensions (academic freedom) not explicitly considered but may influence responses

Limitations

The authors state: "While this study provides an in-depth, comparative analysis of AI regulations in higher education, its findings have limitations in generalizability due to the small sample size of two institutions
Accordingly, interpretations are limited to the participating institutions and are not intended as claims about the entire higher education systems of Türkiye or Czechia
The purposive sampling method may also introduce selection bias, as participants likely had a prior interest in AI governance
Additionally, the reliance on self-reported perceptions of policy agility and effectiveness may not accurately reflect actual implementation." The study also "acknowledges that cultural dimensions, such as academic freedom, were not explicitly considered but may have significantly influenced responses."

Open questions raised

Need for multi-institutional design beyond two case study universities
Integration of explicit cultural factors (academic freedom, institutional autonomy traditions) into analysis
Examination of actual implementation versus self-reported perceptions of policy agility
Longitudinal assessment of how agile governance frameworks perform over time in higher education AI policy
Comparative analysis across more diverse national regulatory contexts
The authors identify the following future research directions: "Future research should therefore utilize a multi-institutional design and integrate cultural factors for a more comprehensive assessment of AI policy in higher education." The study also highlights the need for further investigation into how "cultural dimensions, such as academic freedom," influence AI governance and policy responses in higher education institutions.

Extracted from: pdfAgreement 64%

Explore related topics

Institutional Policy AI Governance Mixed Methods