AI evidence extraction

Automated feedback generation in an intelligent tutoring system for counselor education

Eric Rudolph, Hanna Seer, Carina Mothes, Jens Albrecht · Annals of Computer Science and Information Systems · 2024

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

5/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.40

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.15439/2024f1649

Methodology & findings

Study design

Comparative evaluation study assessing different feedback methods (sandwich, WWW, and STATE methods) in an intelligent tutoring system, with evaluation of emotional and objective impact; includes assessment of AI feedback acceptance among counselor trainees..

Main result

The study found that "AI-generated feedback fulfills objective criteria better than emotional ones" and that "fine-tuning an open source LLM can improve both the emotional and objective quality of feedback." Additionally, the research reveals that "the study examines the acceptance of AI feedback among aspiring counselors, highlighting the influence of familiarity with AI on acceptance levels."

Reports effect sizes.

Research paradigm

Pragmatist/Mixed methods

Author conclusions

The authors conclude that "this research contributes to the understanding of the role of AI in improving digital counseling practices and highlights the need for continuous evaluation and ethical considerations." They further note that "ethical considerations, including bias and hallucination, are addressed, with recommendations for risk mitigation through multi-feedback options and expert supervision."

Risk of bias

Potential AI bias in feedback generation; Hallucination risks in LLM outputs; Selection bias in counselor trainee sample; Potential confounding from participant familiarity with AI systems; Selection bias: Study population limited to aspiring counselors (self-selected sample); Potential confounding: Familiarity with AI may influence both acceptance and performance measures; Potential experimenter bias in feedback evaluation (emotional vs. objective criteria assessment); Limited information on blinding procedures; Potential bias from familiarity with AI influencing acceptance ratings; risk of bias in ethical considerations (bias and hallucination in LLM outputs)

Limitations

The paper addresses limitations through ethical considerations, noting that "bias and hallucination" are potential risks in AI-generated feedback, with recommendations for mitigation through "multi-feedback options and expert supervision."

Open questions raised

The study identifies the need for continuous evaluation of AI-generated feedback in counseling education, further research on ethical considerations including bias and hallucination mitigation, and investigation of how familiarity with AI impacts acceptance among counseling trainees.
The authors identify the need for continuous evaluation of AI-generated feedback in counseling contexts, ethical considerations in AI deployment (bias and hallucination), and further research on acceptance of AI feedback among counselor trainees based on AI familiarity levels.
The study highlights the need for continuous evaluation of AI in digital counseling practices and the importance of addressing ethical considerations in AI implementation for counselor education

Data: not_statedCode: not_statedExtracted from: pdfAgreement 61%

Explore related topics

AI-Assisted Discovery Scholarly Infrastructure Design Science