AI evidence extraction

Fairness, Accountability, Transparency, and Ethics (FATE) in Artificial Intelligence (AI) and higher education: A systematic review

Bahar Memarian, Tenzin Doleck · Computers and Education Artificial Intelligence · 2023

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

7/10

Relevance

1/4

Quality (LMQS)

Evidence

330

Citations

77.30

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1016/j.caeai.2023.100152

Methodology & findings

Study design

Systematic literature review following PRISMA principles.

Sample

N = 33, 3 groups

Primary method

Thematic categorization and descriptive synthesis. Definitions were grouped into descriptive vs. technical categories. Studies were classified as qualitative or quantitative and organized by primary FATE term (Fairness, Accountability, Transparency, or Ethics).

Main result

The systematic review found that "more descriptive definitions exist (especially for fairness) and similarly quantitative definitions mostly emerge for Fairness." Additionally, "more quantitative studies exist (especially for fairness) and qualitative definitions mostly emerge for ethics." The authors also note that "generally, though, there are more definitions than relevant studies conducted in the literature," indicating a significant gap between theoretical definitions of FATE concepts and empirical research implementation in higher education AI contexts.

Reports effect sizes.

Research paradigm

Interpretivist/qualitative synthesis with descriptive analysis

Author conclusions

The authors conclude that "This systematic literature review offers a summary of definitions and studies conducted for FATE terms and AI in higher education literature. Future work may benefit from bridging the gap between laypeople and experts by linking descriptive definitions with technical ones as well as qualitative studies with quantitative ones. Moreover, future work can study accountability and transparency further and make the study of FATE terms more longitudinal, open-access, and reproducible."

Risk of bias

Potential selection bias from database source selection (SCOPUS and WoS only); possible publication bias toward published articles; potential bias in categorization decisions (descriptive vs. technical definitions); Selection bias: Limited to SCOPUS and Web of Science databases, potentially excluding other relevant literature; Publication bias: Systematic reviews inherently capture published literature only; Potential language bias: No mention of language restrictions or inclusion criteria; Definition variability: Subjective categorization of definitions as 'descriptive' vs 'technical'; Selection bias: Limited to SCOPUS and Web of Science databases; may exclude grey literature and other sources; Publication bias: Systematic reviews are susceptible to published study bias; Screening bias: Potential inconsistency in study selection and categorization across reviewers

Limitations

The paper notes that "there are more definitions than relevant studies conducted in the literature," suggesting a limitation in the empirical validation of theoretical frameworks
The abstract also indicates that future work should "bridge the gap between laypeople and experts by linking descriptive definitions with technical ones as well as qualitative studies with quantitative ones," implying current limitations in integrating different definitional and methodological approaches.

Open questions raised

The authors identify the following gaps: (1) a gap between laypeople and expert definitions that needs bridging; (2) disconnect between descriptive and technical definitions; (3) mismatch between qualitative and quantitative study types; (4) need for more studies on accountability and transparency; (5) need for more longitudinal, open-access, and reproducible FATE research
Gap between laypeople and expert definitions of FATE concepts
Mismatch between qualitative and quantitative studies
Limited research on accountability and transparency compared to fairness
Need for more longitudinal studies
Need for open-access and reproducible FATE research

Data: not_statedCode: not_statedExtracted from: pdfAgreement 64%

Explore related topics

Responsible AI Use AI Governance Literature Review