Task Spec: t-diversity-bonus — Diversity Bonus in Resource Allocation

Task Spec: t-diversity-bonus — Diversity Bonus in Resource Allocation

Goal

Modify compute_allocation_weight() in exchange.py to include a diversity_bonus factor that rewards hypotheses targeting underexplored mechanisms.

Approach

Add a _compute_diversity_bonus() helper function that:

- Counts existing hypotheses targeting the same gene and pathway
- Returns a bonus multiplier in range 0.4–1.2
- Unexplored mechanisms (0 existing) get 1.2x bonus
- Heavily studied mechanisms (6+ existing) get 0.4x penalty

Update compute_allocation_weight() to multiply in the diversity bonus:

- Formula: allocation_weight = composite_score evidence_freshness gap_coverage * diversity_bonus
- Bonus schedule:
- 0 existing hypotheses: 1.2 (completely unexplored)
- 1 existing: 1.1 (barely explored)
- 2-3 existing: 1.0 (moderately explored, neutral)
- 4-5 existing: 0.8 (well-studied)
- 6+ existing: 0.4 (heavily explored, deprioritized)

Acceptance Criteria

Work Log

• 2026-04-20 12:35Z: Re-implemented after task reopen audit. The original commit 41579dcc6 was lost during squash merge 1fe00aba0. Re-applied _compute_diversity_bonus() and integrated it into compute_allocation_weight(). Committed as 5dcf7f478.
• 2026-04-17: Implemented _compute_diversity_bonus() helper function and integrated it into compute_allocation_weight(). The diversity bonus multiplier (0.4–1.2) is applied as a 4th factor in the allocation weight formula, rewarding hypotheses targeting underexplored mechanisms and penalizing those targeting already-heavily-studied areas.

Tasks using this spec (1)

Diversity bonus in resource allocation

File: t_diversity_bonus_spec.md

Modified: 2026-04-24 07:15

Size: 2.0 KB