Context‑Sensitive AI Chatbots for Scalable Teacher Professional Development: A Mixed‑Methods Evaluation in Austrian Higher Education

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: Division K - Section 06: In-Service Teacher Professional Development Virtual Poster Session

Abstract

Continuous professional development (PD) for in‑service teachers is frequently misaligned with individual needs, resulting in low uptake and limited instructional impact. Leveraging a Retrieval‑Augmented Generation chatbot, we delivered context‑sensitive PD recommendations to 4 267 teachers in Burgenland (Austria) over six weeks. From 2 030 valid interactions (n = 1 125 teachers) we extracted quantitative key‑performance indicators, fallback rate, intent interpretation, sentiment and confidence, and triangulated them with qualitative content analysis of open comments in a convergent‑parallel mixed‑methods design. Integrating the Information Systems Success Model (ISSM) with TAM and TPACK, we report a fallback rate of 14.4 % (below the 20 % threshold) and 85 % positive sentiment, evidencing robust system and information quality. Query specificity (ρ = .36, p < .05) and TPACK‑aligned keywords (OR = 2.05) significantly predicted user satisfaction. Time‑series peaks coincided with institutional PD announcements, demonstrating organisational triggers for adoption. Findings inform design guidelines for evidence‑based, GDPR‑compliant teacher‑facing chatbots and delineate research avenues on long‑term classroom impact.

Authors

• Thomas Leitgeb, University of Education Burgenland

• Michael Leitgeb, University College Of Teacher Education Burgenland