EXPLAINING THE EFFICACY OF THE BRCA GIST INTELLIGENT TUTOR FOR GENETIC TESTING AND BREAST CANCER RISK
Purpose: Previous experiments demonstrated the efficacy of the BRCA Gist Intelligent Tutoring System (ITS) in helping women understand and make decisions about genetic testing for breast-cancer risk. The purpose of this experiment was to understand which aspects of BRCA Gist are responsible for particular outcomes, shedding light on cognitive mechanisms.
Methods: BRCA Gist avatars presented information orally, in writing, and through graphics and video clips designed following Fuzzy-Trace Theory (FTT). Interactive dialogues between user and avatar addressed questions such as, "How do genes affect breast-cancer risk?" Using "expectations texts" and Latent Semantic Analysis, BRCA Gist "understands" and responds to participants' typed statements helping them form good gist explanations of five aspects of BRCA1/2 testing. Women (N=206) were randomly assigned to BRCA Gist, a control ITS, or selectively reduced versions of BRCA Gist. The No-Tutorial-Dialogues condition was the same as BRCA Gist except it omitted dialogues that guide gist explanations. The No-FTT-Images condition was the same as BRCA Gist except 10 images designed following FTT that help users form gist representations (example below) were removed. Theory-guided images included figures that communicated base rates of genetic mutation and of breast cancer, ordinal gist of survival rates at different stages, and conditional probabilities (e.g., of breast cancer given BRCA mutation). We assessed declarative knowledge, gist comprehension, and genetic decision-making. Dialogues were subjected to fine-grained analyses of gist-level understanding.
Results: Participants in all BRCA Gist conditions scored significantly higher on key outcomes, demonstrating the efficacy of BRCA Gist. BRCA Gist scored higher than both No-Tutorial Dialogues and No-FTT-Images groups on two tests of declarative knowledge. Detailed analyses indicate that removing dialogues and FTT images reduced performance on knowledge of mutations and genetic testing. Regarding genetic testing decisions, generating explanations and FTT images increased the weight participants gave to the number of relatives with cancer, improving concordance with testing recommendations. Generating explanations also improved genetic risk categorization. Fine-grained content analysis of verbal responses in dialogues further explicated underlying mechanisms.
Conclusions: Evidence of BRCA Gist efficacy replicated key findings from laboratory, web, and community studies, consistent with mechanisms of FTT. Images created using FTT principles and generating gist explanations each appears to enhance mental representations of base rates of BRCA mutations and breast cancer, recommendations for testing, and consequences of testing.