Talk it through: conversational AI for sexual consent education

Thesis

Consent education in universities has become a legal and ethical necessity. Typically delivered through static e-learning formats, current trainings often result in low engagement, short-lived effects and minimal behavioral impact. This thesis evaluates whether a large language model chatbot can deliver more effective consent and sexual harassment training than the standard e-learning module used at UK universities. Four outcomes are examined: user reaction, cognitive learning, affective learning and behavior. A two-arm randomized controlled trial assigned 150 UK-based students to either a control e-learning module (n= 76) or an equivalent chatbot training (n= 74) for treatment, with outcomes measured via pre-training, post-training, and two-week follow-up questionnaires. Both formats were similarly well-received (UX_t = 8.7,UX_c = 8.84). Knowledge increased in each group with a higher effect from the chatbot (η² = 0.03). Attitude scores improved similarly in both groups. Behavioral intention to pursue further training was higher in the treatment group (d = 0.29). The chatbot training matches or exceeds the performance from traditional methods, offering a promising, accessible, private and low-cost complement to in-person consent trainings.

Authors: Gutenmacher, S

• DOI: 10.5287/ora-6g9njjzgb
• Type of award: MSc taught course
• Level of award: Masters
• Awarding institution: University of Oxford
• Language: English
• Keywords: sexual consent education; large language models
• Subjects: Artificial intelligence--Educational applications; Sexual harassment--Prevention