Kevin D. Ashley

Evaluates a novel instrument for assessing moral reasoning skills in bioengineering ethics education, comparing it to existing techniques and testing validity and reliability. Findings show the instrument is sensitive to knowledge gains and that independent coders can reliably apply it, supporting its use for assessing complex ethical case analysis.

Compares two types of rating prompts in computer‑supported peer review—domain‑writing composition prompts versus problem‑specific prompts—to examine effects on reviewer sensitivity, correlation with instructor scores, and informativeness for authors. Results indicate reviewers distinguish problem‑specific issues better, both prompt types correlate with instructor scores, and problem‑specific ratings yield more helpful and less redundant feedback to peer authors.

Presents SMILE+IBP, which bridges case‑based reasoning and text information extraction to predict outcomes from case texts. The system classifies fact descriptions using Factors—stereotypical fact patterns affecting claim strength—and evaluates and explains predictions against a database of previously classified cases. The paper illustrates prediction by IBP and text classification by SMILE and reports empirical evaluations of both functions in the trade secret domain.

Reexamines what makes problems and domains “ill‑defined” for AI in Education. Proposes definitions focusing on when essential concepts, relations, or criteria are under‑specified, open‑textured, or intractable, requiring problem recharacterization. The framework aligns structural and pedagogical features of ill‑definedness to support research and design of AIED systems and tasks.

Introduces LARGO, an intelligent tutoring system that helps students learn legal reasoning with hypotheticals by diagramming U.S. Supreme Court oral arguments. Students use a specialized graphical language and receive feedback via reflection questions. The paper describes targeted reasoning skills (proposing tests, posing and responding to hypotheticals, analogizing and distinguishing) and reports initial evaluation results.

Compares computerized algorithms for predicting legal outcomes not only on accuracy but also on their ability to explain predictions and integrate arguments. Describes the Issue‑Based Prediction (IBP) algorithm, which tests hypotheses about issue decisions, seeks to explain away counterexamples while apprising users of them, and generates explanatory arguments from case comparisons.

HYPO is a case‑based reasoning system that evaluates legal problems by comparing and contrasting them with cases in its knowledge base, generating adversarial arguments and posing hypotheticals that shift the evaluation. Using dimensions to access and assess cases, HYPO computes relevant similarities and differences, identifies best cases to cite, generates counterexamples, and selects targets for hypotheticals. These definitions support context‑sensitive assessments of relevance and salience without requiring strong domain theories or a priori weights.

Overviews HYPO, a case‑based reasoning program operating in trade secret law. The paper identifies key ingredients of CBR, relates them to HYPO’s mechanisms, and demonstrates HYPO working through a hypothetical trade secrets case patterned after an actual dispute, illustrating retrieval, analysis, and argument construction with precedents.