The Effect of Displaying System Confidence Information on the Usage of Autonomous Systems for ...
summary
A lab study with 60 participants tests a Wizard-of-Oz OCR agent against a manual grammar task under a 30-minute/30-task limit and a £10 payout cap. The UI shows five confidence levels (color-coded) or hides them; confidence is simulated (Wizard of Oz) and still contains errors. Displaying confidence increases agent task completion and blind acceptance significantly, and participants adopt varied strategies and interpretations rather than uniformly trusting or rejecting the autonomous system.
AI-written summary · about
chapters (7)
- 0:00Introduction and motivation
- 3:33Study design and methods
- 8:11Task interface and conditions
- 10:17Quantitative findings
- 12:25Qualitative insights and strategies
- 14:13Summary and implications
- 15:29Q&A session
related talks
45mBeing Smart vs. Feeling Smart
25mSingle or Multiple Conversational Agents? An Interactional Coherence Comparison
17mExploring the Role of Conversational Cues in Guided Task Support with Virtual Assistants
09mBuilding Trust in Conversational AI
11mRuntime Permissions for Privacy in Proactive Intelligent Assistants
01h 04mWhat are we thinking in the age of AI?
16mTrust by Design: Building Reliable AI for Sensitive Environments
12mThe Comfort Trap: Getting Too Used to AI