Technology Dominance in Complex Decision Making: The Case of Aided Credibility Assessment

Abstract

Decision aids have long been an important source of help in making structured decisions. However, decision support for more complex problems has been much more difficult to create. Decision aids are now being developed for very complex problems, and their effects among low- and high-task-knowledge individuals are still being explored. One such task is credibility assessment, in which message recipients or observers must determine a message's veracity and trustworthiness. Credibility assessment is made difficult by lack of constraints, hidden or incomplete information, and mistaken beliefs of the assessor. The theory of technology dominance (TTD) proposes that technology is most effectively applied in intelligent decision aids when an experienced user is paired with a sophisticated decision aid. This work examines TT D in the complex task of credibility assessment. To assist in credibility assessment, we created a decision aid that augments the capabilities of the user—whether novice or professional. Using hypotheses based on TT D, we tested the decision aid using high-stakes deception in recorded interviews and involved both student (novice) and law enforcement (professional) users. Both professionals and novices improved their assessment accuracy by using the decision aid. Consistent with TTD, novices were more reliant on the decision aid than were professionals. However, contrary to TTD, there was no significant difference in the way novices and professionals interacted with the system, and the decision aid was not more beneficial to professionals. Novices and professionals frequently discounted the aid's recommendations, and in many cases professionals did not view explanations when the decision aid contradicted their assessments. Potential reasons for these findings, as well as limitations and future research opportunities, are discussed.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 82 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.