## 1. probabilsitic program
1. Probabilistic inference provides a robust methodology for dealing with dynamic and multi-stakeholder complexities in entrepreneurial decisions. Traditional optimization methods (e.g., integer programming, branch-and-bound) lack flexibility under rapidly changing entrepreneurial conditions, whereas probabilistic models handle adaptive learning more naturally, allowing entrepreneurs to dynamically adjust their decisions based on real-time feedback
## 2. group prior
[[tan_zhixuan]]'s recommendation on social prior from [[session7_social planner]] + (Inverse) planning coordination systems
- [[📜planning with theory of mind for few shot adaptation in sequential social dilemmas]]
## 3.pomdp - lp
- POMDP literature on information gathering, myopic policies, LP approximations for MDPs
# 4. 📦multiple hypothesis testing + inventory management
**Markovian Prompt: Supply Chain Math for Startup Testing Decisions**
**Main Mathematical Framework:**
1. **Basic Testing Decision Equation**:
- <span style="color: blue">ΔEU(n)</span> = <span style="color: green">[nα/(α+n)]</span><span style="color: red">(μ-φ_true)</span> - <span style="color: orange">nc^y</span> + <span style="color: purple">c^φ</span>
- Where:
- n = number of tests (e.g., cars to test)
- α = prior confidence
- μ = belief about market success
- φ_true = actual market reality
- c^y = cost per test
- c^φ = fixed cost of alternative research
2. **Multi-Stakeholder Decision Rule**: Test stakeholder j if: <span style="color: green">w_j·InfoGain_j</span> + <span style="color: blue">Σ_{i≠j} w_i·Spillover_{j→i}</span> > <span style="color: red">γc_j</span>
**Testing Errors Meet Inventory Theory Table:**
|Testing Concept|Inventory Theory Equivalent|Startup Application|TAXIE Example|
|---|---|---|---|
|**Type I Error (False Positive)**|**Overage Cost**|Launching bad product|Building a large EV fleet when market isn't ready|
|• Probability: α|• Excess inventory risk|• Wrongly entering market|• Risk of unsold car capacity|
|• Cost: C₁|• Cost of unsold goods|• Wasted investment|• Cost of idle EVs|
|**Type II Error (False Negative)**|**Underage Cost**|Missing good opportunity|Abandoning EV rideshare too early|
|• Probability: β|• Stockout risk|• Skipping winning idea|• Missing first-mover advantage|
|• Cost: C₂|• Lost sales cost|• Competitor wins market|• Tesla/others capture market|
|**Prior Belief (μ)**|**Demand Forecast**|Market size belief|TAXIE believed high demand for EV rideshare|
|**Prior Strength (α)**|**Forecast Confidence**|How certain you are|Low confidence (needed testing)|
|**Optimal Sample Size**|**Economic Order Quantity**|How many to test|n* = 2-3 cars was optimal|
|**Total Cost Minimization**|**Total Inventory Cost**|Balance all error costs|Minimize false launches + missed opportunities|
**Easy Explanation:**
Think of testing like ordering inventory:
- Order too much (Type I error) = Launch product nobody wants
- Order too little (Type II error) = Miss a great opportunity
- The goal: Find the sweet spot that minimizes total mistakes
**TAXIE Case Study:**
TAXIE tested with 2 cars and learned:
1. **Range hypothesis**: ✓ Confirmed (260 miles sufficient)
2. **Driver earnings**: ✓ Validated initially
3. **Willingness to pay**: ✓ $400/week seemed viable
4. **Infrastructure**: ? Needed larger test (>50 cars)
5. **Scalability**: ? Systems untested at scale
6. **Profitability**: ✗ Not viable at current scale
**Optimal Sample Size Calculation:** Using the formula n* = √[α²(μ-φ_true)/c^y] - α
For TAXIE:
- Prior belief (μ): 0.5 (moderate optimism)
- True reality (φ_true): 0.2 (market was actually weak)
- Prior confidence (α): 2 (low confidence)
- Cost per car test (c^y): 0.15
- Spillover factor: 1.5 (testing reveals multiple insights)
n* = √[4 × 0.3 × 2.5 / 0.15] - 2 = √[20] - 2 = 2.47 ≈ 2-3 cars
**Key Insights from TAXIE:**
1. Testing 2 cars was mathematically optimal given their constraints
2. They correctly learned about range, driver economics, and pricing
3. They couldn't test infrastructure and scalability with just 2 cars
4. The spillover effect (one test reveals multiple insights) justified the investment
**Visual Requirements:**
1. **Decision Flowchart**: Show how TAXIE moved through hypothesis testing
2. **Learning Curve**: Plot information gained vs. number of cars tested
3. **Error Cost Diagram**: Visualize Type I vs Type II error tradeoffs
4. **Spillover Network**: Show how testing drivers revealed insights about infrastructure, pricing, and operations
**Main Takeaway:** Supply chain math helps startups test efficiently. Just as inventory managers balance overstocking vs. stockouts, entrepreneurs must balance over-investing in bad ideas vs. missing good opportunities. The optimal testing strategy minimizes total error costs while maximizing learning through spillover effects.
---
# relevant papers from [[📝moon24_csv_ai_cofounder]]
| Paper Title | Reason for Classification (Hypothesis Testing & Stakeholder Complexity) | Optimization Component (3.1 Theoretical) |
| ------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Adaptive Entrepreneurship: A Preliminary Framework Using Exaptation and Exchangeability** | Uses exaptation and cross-context hypothesis testing to proactively explore and validate stakeholder opportunities, strategically reducing uncertainty across markets and stakeholders . | Formulates stakeholder uncertainty reduction via exchangeable decises, systematically structuring state-action hypotheses for proactive testing ($U_d$) . |
| **Zero to One and Done** | Systematically classifies and tests uncertainties across contexts (epistemic and aleatoric) to proactively generalize validated insightsple stakeholder opportunities, thus strategically reducing complexity . | Clarifies how proactive hypothesis testing accelerates uncertainty reduction through validation of exchangeability across contexts, refining stakeholder-focused state-action-utility relationships ($U_d$) . |
| **Equity Proposal as Action: Optimal Term Sheets via Conversational Inference** | Implements iterative hypothesis-probing proposals during negotiations, reducing stakeholder complexity through structured conversational inference about investor intentions and founder-investor alignments . | Defines proactive hypothesis-driven negotiation strategies as structured sequences of proposals, explicitly mapping stakeholder state-action-utility interactions ($U_i$) . |