Biodiversity & Discovery · Week 7 of 9
Can we trust the machine?
Test it · Judge it
FORGE Test · Evaluate
Today's goal
Test MaramaTrap's AI identifications against our own and an expert's, and evaluate its confidence.
By the end
Audit table complete with a reason for each disagreement.
What we'll do (60 minutes)
- What does 'confidence' mean for an AI?10 min
- Audit: our ID, AI ID and confidence, expert ID, agree?25 min
- Evaluate: when is high confidence wrong? when is low right?15 min
- Write the class trust rule10 min
Need a hand?
5 photos with the expert answers given.
Want a challenge?
Test photo quality against AI confidence.
Try this week
AI this week
MaramaTrap moth identifier: Suggests which moth family is in a photo and says how confident it is (from 0 to 1). You: Check it against your own ID and an expert's. Decide when to trust it and when not to.