Does a likelihood-ranked multi-start make the hardness-aware EM robust to a poor start?

What it found

Fitting the three-auditor hardness-aware Dawid–Skene model from 8 grid starts (none near the truth) and keeping the highest-likelihood fit recovers the hard share at the population level (P(hard) 0.300), where a single poor start stays at 0.500 with a lower likelihood. At finite n it matches one informed start (P(hard) 0.290 vs 0.290 at n = 1000, hard accuracy 1/3) and avoids the poor start’s utility loss (0.3913 vs 0.3843). With no hardness at all the hard share is not identified, and fitted shares of 0.3–0.4 appear anyway. Stylised.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Code
• Tests

More in Ambient AI assistant

• How many starts does the likelihood-ranked multi-start need, and are grid starts better than random ones?How many starts does a likelihood-ranked multi-start need, and are grid starts better than random ones: one start is nearly enough with strong hardness or n = 1000, 4 starts reach the best optimum in most cells, more help only in the weak-hardness and null cells, and policy utility barely moves (stylised)
• Can a likelihood-ratio test tell “no hard messages” from “some”, and is the chi-square cutoff usable?Can a likelihood-ratio test tell no hard messages from some: chi-square cutoffs reject 35-72% of null data sets, a parametric bootstrap holds its level and detects h = 0.3 at n = 1000 but not h = 0.1 (stylised).
• Does the bootstrap likelihood-ratio test of “no hard messages” hold its level?Does the bootstrap likelihood-ratio test of no hard messages hold its level: 200 null draws per n give rejection 0.040-0.055 at nominal 0.05, all intervals containing 0.05 (stylised).
• Does the bootstrap test of “no hard messages” stay valid when the auditors differ in accuracy?Does the no-hard-messages bootstrap test keep its level when the three auditors differ in accuracy: 100 h = 0 draws per cell give rejection 0.00-0.10 at nominal 0.05, no inflation seen (stylised).
• Can an always-on assistant learn when to speak from accept/dismiss feedback alone, and what does it cost in dismissals?Can an always-on assistant learn when to speak from accept/dismiss feedback alone: greedy, epsilon-greedy and Thompson online policies vs an offline oracle, mostly negative (stylised)
• Does inverse-propensity weighting fix an always-on assistant learning from its own accept/dismiss feedback?Does inverse-propensity weighting fix an assistant learning from its own accept/dismiss feedback: mostly negative, exploration cost dominates (stylised)

All 46 Ambient AI assistant projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.