Interactive Model Demo

Conversation Risk Scorer

Input session characteristics to predict failure risk and mode — powered by a nested random forest trained on 1,250 AI assistant conversations.

0% FAILURE RISK Low High
Session Parameters
Task Type
Primary Intent
Total Turns 8

Number of exchanges in the session (2–19)

Avg Sentiment 0.00

Mean user sentiment across turns (−1.4 negative → +1.7 positive)

Sentiment Range 2.68

Emotional volatility — max minus min sentiment (0–5.6)

Intent Confidence 0.78

Avg bot confidence in intent recognition (0.58–0.94)

Avg Latency (ms) 478

Mean bot response time in milliseconds (120–7004)

Total Failures 2

Turns where resolution was unsuccessful (0–7)

Risk Assessment
Level 1 — Outcome
Confidence:
Level 2 — If Failed
Confidence:
Behavioral Signal Strength
Recommended Intervention
Dataset: Mohsen Rafiei — UX Datasets (GitHub)  ·  Model: Nested Random Forest, scikit-learn  ·  Built by Ishika Ray