Work

Traceable Texts and Their Effects: A Study of Summary-Source Links in AI-Generated Summaries

Year: 2025

Type: conference-paper

Abstract: AI-generated summaryThe patient shows slight tiredness, somewhat high blood pressure, a gentle heart murmur, and mildly swollen neck glands, but otherwise has a normal exam.He's scheduled to come back in two months for another follow-up to review his progress.Physical Exam: Patient appears fatigued.Weight: 179 pounds.Blood Pressure: 142/80; 136/78.Neck: Mild cervical lymphadenopathy without thyromegaly.Lungs: Clear.Cor: S1, S2 normal without gallops or rubs, II/VI systolic ejection murmur at the apex.Abdomen: Non-tender, without organomegaly.Extremities: No edema.Figure 1: Interacting with a traceable text.We call an AI-generated summary with backlinks to a source text a "traceable text." Pictured is the kind of traceable text we studied in this paper.The generated summary is augmented with phrase-level links (e.g., "mildly swollen neck glands") to corresponding passages in a source text (e.g., "Neck: Mild cervical lymphadenopathy without thyromegaly").The links help a reader to check the veracity of a passage in a summary and also offer an index into the source text if the reader wishes to learn more.

Source: Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA)

DOI: 10.1145/3706599.3719830

Authors: Hita Kambhamettu, Jamie Flores, Andrew Head

Institution: University of Pennsylvania

Language: English

Topic: Text Readability and Simplification

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: bronze

Funders: U.S. Department of Defense, Defense Advanced Research Projects Agency, Advanced Research Projects Agency

OpenAlex ID: W4409906283

FWCI: 3.964

Cites: 39

Cited by: 4

Related to: 10