Special-issue evidence: a knowledge-graph-enriched legal-writing textbook led to better learning outcomes than baselines, and aggregated emoji annotations helped teachers identify challenging text¶

Claim · All claims
Evidence · 1 study (2 entries) · review r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Sovrano et al. (2024) report an experiment in legal writing education in which the knowledge-graph-enriched textbook led to better learning outcomes than the baselines. → Sergey Sosnovsky 2025 q2 i? Blobstein et al. (2024) demonstrate that aggregating students' emoji/hashtag annotation data could help teachers understand which parts of the text have been challenging or interesting. → Sergey Sosnovsky 2025 (2)

Evidence¶

Sergey Sosnovsky 2025¶

Sergey Sosnovsky, Peter Brusilovsky, Andrew Lan. (2025). Intelligent Textbooks. International Journal of Artificial Intelligence in Education. https://doi.org/10.1007/s40593-024-00451-9

q2 · i? · review · r2

The editorial summarizes the special-issue study by Sovrano et al. (2024), which extracts a knowledge graph from a collection of textbooks and other learning resources to provide on-demand explanations, combining knowledge graphs with generative question-answering models. No effect size is printed in the editorial.

"The authors conducted an experiment in the context of legal writing education to show that the enriched textbook leads to better learning outcomes than the baselines"

Sergey Sosnovsky 2025 (2)¶

Sergey Sosnovsky, Peter Brusilovsky, Andrew Lan. (2025). Intelligent Textbooks. International Journal of Artificial Intelligence in Education. https://doi.org/10.1007/s40593-024-00451-9

q2 · i? · review · r2

The editorial summarizes Blobstein et al. (2024), which leverages students' emoji and hashtag annotations of course readings as a fourth-generation interaction-data source and also explored the feasibility of using LLMs to predict student annotations.

"The paper demonstrated that aggregating student annotation data could help teachers understand which parts of the text have been challenging or interesting"

Discussion¶

Related Claims¶

• The ten special-issue papers collectively indicate near-term feasibility of automated guidance and feedback on reading and writing skills — a broader claim this one bears on
• Explanations generated from generic questions were rated significantly better than those generated from random questions, supporting the hypothesis that question selection strategy matters — related
• A DoX-based pre-filtering step reduces the Intelligent Explanation Generator's asymptotic time complexity from quadratic to linear in the size of the text collection — related
• In a within-subjects user study with more than 100 students, explanations generated by the Intelligent Explanation Generator were rated significantly higher than those of random-question and generic-question baselines — related
• The fine-tuned T5 question-extraction model reached an average training loss of 0.4098, indicating imperfect question extraction — related
• Cognitive and metacognitive prompts in learning journals helped early in a term but became detrimental by its end; adaptive fading-out mitigated this — related