2025-03-12 reacting to [[charlie_fine]]'s need on venturing in transportation , i applied theories from below to [[Space/School/24_🚗Zhao23_behav_mob_ai/🛝mmi24_2]], [[MMI forum]], [[mmi2024]] for my [[⭐️thesis]] [[3.1 Entrepreneurs are not secretaries]] [[3.2 bayesian_calibration]] [[3.3 bayesian calibrated choice - balancing individual and market-level uncertainty]] [[📜⏰mendelson99_Industry Clockspeed Measurement and Operational Implications]] [[Pasted image 20250118221138.png|600]] | | Title with abstract | Sections | fig | Page | | ----------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------- | ---- | | [[🌓⌨️Entrepreneurs Are Not Secretaries_Decision Making when State and Action are Indistinguishable]] | | | | | | | Complexity of Entrepreneurial Decision Making and Strategy<br><br>1Tesla’s example<br><br>EDM’s components are x t (decision), R t (ratio performance measure), c t (fixed costs), At (constraint matrix), b t (constraint limits), U t (Utility function), L (Utility lower bound). EDM’s hyperparameters are N is the number of decision variables in each time period. M is the number of constraints in each time period. T is the number of time periods over which decisions are made. P is the number of performance measures or outcomes affected by the decisions. Q is the number of possible opportunity states or scenarios.<br><br>Table 1 illustrates Tesla’s decision-making process using the EDMNO framework, comparing value creation and value capture decisions. In both scenarios, binary decision variables (x t = [x 1t , x 2t ], where N = 2) represent different choices: for value creation, they indicate investments in battery technology or marketing, while for value capture, they represent make-or-buy decisions for battery packs and power electronics modules (PEM). The outcomes of these decisions are represented by intermediary variables. In value creation, f 1 (x 1t , t) and f 2 (x 2t , t) correspond to battery performance and vehicle sales, respectively. For value capture, g 1 (x 1t , t) and g 2 (x 2t , t) represent battery pack and PEM quality. These are approximated as w it x it , where w it are time-dependent weights. The utility functions (U t ) for both scenarios are weighted sums of these performance measures minus associated costs. The weights (W t ) in the utility function reflect the relative importance of each performance measure. Fixed costs (c t (x t )) vary by year for value creation and depend on make-or-buy decisions for value capture. Constraints are represented by matrix At and vector b t . For value creation, these include cost and resource requirements of each investment, along with budget and resource availability over time. For value capture, they encompass resource requirements for in-house production versus supplier capacity, as well as production capacity, quality thresholds, and budget constraints. | - Tesla's example<br>- Decomposing Complexity<br>- Entrepreneurial Strategy across Lifecycle<br>- Formal definition of EDM, EDMNO, ILP, Knapsack<br>- Formal Proof: Reduction from 0-1 KNAPSACK to EDM<br>- P-NP of EDM | ![[Pasted image 20241124075000.png\|200]] | 120 | | | Integrative Models of Entrepreneurial Pivoting: A Simulation Approach | | | 1 | | [[🌒📐Test quantities shape sensitivity]] | | | | | | | Rationalizing Entrepreneurial Learning<br><br>Over the past decades, entrepreneurial learning has been examined through diverse lenses—scientific, Bayesian, evolutionary, environmental, behavioral, effectual, to name a few. While these approaches have yielded valuable insights, there were limited integration efforts. This paper propose a framework that can be uniformly applied to rationalize different approaches of entrepreneurial learning.<br><br>We categorize existing entrepreneurial learning concepts into three primary approaches—Behavioral, Bayesian, and Evolutionary—and draw parallels with five established machine learning paradigms: Symbolist, Bayesian, Evolutionary, Connectionist, and Analogizer. This allows us to compare entrepreneurial learning mechanisms by their representation, evaluation, and optimization algorithms. To formalize these approaches, we apply rational analysis by defining abstract computational problem each system (agent) solves and explain its behavior in terms of optimal solutions. | - Introduction<br>- Rational Analysis and Entrepreneurial Strategy<br>- Entrepreneurial and Machine Learning<br>- Mapping Entrepreneurial to Machine Learning<br>- Rational Analysis of Entrepreneurial Learning<br>- Predicting Convergence<br>- Conclusion<br>- Appendix | ![[Pasted image 20241124072847.png\|200]] | 38 | | | AI Co-founder for Pivoting Decisions<br><br>We present a Human-AI collaboration framework to address challenges in entrepreneurial decision-making, where each action both informs and alters the business environment. The high-dimensional and rapidly evolving nature of this environment underscores the entrepreneur’s need to make strategic decisions with incomplete information. We formalize this entrepreneur’s adaptation as pivoting, defined as the change of optimal action in response to updated world models. This assumes entrepreneur’s rationality given infinite computational capacity. Scoping the problem as ”entrepreneurs rarely pivot with computational cognitive decision support”, we propose three-step solution: (1) we interpret entrepreneurial axioms with components of rational agency equation. Then we decompose the equation by functions, situating them in entrepreneurial environments that are dynamic and hierarchical. Using Bayesian software, we implement computational cognitive engine to support pivoting decisions with data pre-processing, parameter management, and scenario generation. (2) we apply computational cognitive engine with world model on Tesla Roadster’s product and supply chain development case. (3) we identify communities that can contribute to productizing this research as ”business intelligence supporting pivot decisions based on a world model”. This integrated approach provides a new paradigm for entrepreneurial decision-making in complex, high-stakes environments, potentially increasing the success rate of innovative ventures. | - (Supply) Need for Automated Decision Support Systems<br>- (Process) Integration of AI in Strategic Planning<br>- (Demand) Founder-AI Collaboration Framework | ![[Pasted image 20241124072111.png\|200]] | 4 | | | AI Co-founder: A Computational Cognitive Approach to Entrepreneurial Paradox | 1. Introduction<br>2. AI Equation and Entrepreneurial Axioms<br>3. AI Equation and Function<br>4. Applying to Tesla's strategy<br>5. Emerging Desirability and Feasibility<br>6. Conclusion and Future Directions<br>7. Appendix | ![[Pasted image 20241124073905.png\|200]] | 80 | | [[🌔]] | | | | | | | Finding Machine's Use for Entrepreneurial Choice | - Choice set, group, triplet<br>- Use of simulation<br>- Entrepreneur with Machine as a Tool<br>- Conditions where simulation is helpful | ![[Pasted image 20241124074747.png\|200]] | 114 | | | Equity Proposal as Action Converging towards Optimal Term Sheets with Conversational Inference<br><br>Abstract: This paper introduces a prototype toolbox for startup pivoting, employing a cognitive system’s multi-level framework to reconcile tension between action and optimization in entrepreneurship. By modeling founders and investors as resource-rational agents, we interpret inference-based proposals as actions, viewing term sheet negotiations as a convergence process toward optimal solutions. In this context, conversational inference between parties represents an information processing mechanism aimed at maximizing expected utility given shared beliefs. Our approach synthesizes entrepreneurial strategy, Bayesian decision theory, probabilistic programming, and conversational inference across computational, algorithmic, and implementation levels. This toolbox facilitates counterfactual reasoning, scenario planning, and strategic choice testing, bridging the gap between abstract optimization objectives and concrete, iterative actions in uncertain entrepreneurial environments. | - Introduction<br>- Literature Review<br>- Three level architecture and implementation<br>- Theory, Algorithm, Implementation for Equity Valuation | ![[Pasted image 20241124075439.png\|200]] | 126 | | | Abstract for Conversational Termsheet Advisor<br><br>Abstract for conversational termsheet advisor: During entrepreneur’s capitalization, value is created by the act of collaborative valuation and captured by competitive control. We use a term sheet negotiation situation between founder and investor to illustrate this idea and show how probabilistic inference drives these actions. We implement conversational termsheet advisor and introduce usecases in 0.1 then explain the meaning of this tool in entrepreneurship context in 0.2 and 0.3.<br><br>Abstract: We illustrate how new observation updates the answer to query for specific decisions in term sheet and how context (industry and company stage) represented as world model affects statistical and pragmatic meaning of observation. | - Three usecases of termsheet advisor<br>- Modeling quality in entrepreneurship<br>- Effect of unifying statistical and pragmatic linguistics | ![[Pasted image 20241124075543.png\|200]] | 138 | | | “Adaptive Entrepreneurship: A Preliminary Framework Using Exaptation and Exchangeability”<br>How does an agent’s adaptive system evolve as its environment shifts from simple to complex? Drawing inspiration from evolutionary theories, we introduce exaptation—a lesser-known evolutionary mechanism—and frame it within a Bayesian perspective using the concept of exchangeability. We identify three distinct innovation pathways: direct adaptation, co-opted adaptation, and coopted nonaptation. Each pathway is defined by unique decision processes and probability structures, ranging from fully dependent (non-exchangeable) sequences to partially or fully exchangeable configurations that allow for unexpected functional shifts.<br><br>Applying this framework, we link horizons of opportunity with hierarchical learning to guide strategic decision-making under different types of uncertainty. Case studies—including Tesla’s Powerwall, BYD’s Blade Battery, and the historical Ford assembly line—illustrate how these evolutionary-inspired dynamics unfold in technological, biological, and business contexts. By managing aleatoric and epistemic uncertainty through exchangeability-based strategies, we prescribe design patterns that balance incremental improvements with bold exploration. In doing so, we show how nature’s evolutionary principles can inform innovation, enabling us to emulate its logic without directly replicating its underlying biological implementations. | | | | | | Zero to One and Done<br>How do entrepreneurs navigate uncertainty in innovation? This paper introduces three key contributions to entrepreneurial decisionmaking under uncertainty. First, we develop a novel epistemic-to-aleatoric (E/A) ratio framework that quantitatively guides exploration strategy as uncertainty evolves. We show how successful exploration of high-uncertainty domains (where both solution and need are new) systematically creates opportunities for medium-uncertainty expansion (where either solution or need is proven). Second, we prove that entrepreneurs can accelerate this evolution through systematic verification of exchangeability—whether success in one context predicts success in others. When entrepreneurs identify such patterns, they can transform initial high-uncertainty experiments into medium-uncertainty strategic options, as exemplified by how platform companies leverage validated technologies across new applications. Third, we provide a dynamic classification mapping uncertainty levels to optimal strategies: high uncertainty (p ≈ 0.6) requires parallel exploration to validate fundamental assumptions, medium uncertainty (p ≈ 0.75) enables iterative sampling of proven components, and low uncertainty (p ≈ 0.9) allows quick optimization decisions. Using hierarchical Bayesian analysis and Monte Carlo simulations, we demonstrate how the E/A ratio systematically evolves across these uncertainty classes, explaining why successful high-uncertainty exploration often spawns multiple medium-uncertainty opportunities. This framework helps entrepreneurs structure their innovation journey: starting with carefully chosen parallel paths in truly novel domains, then systematically leveraging that learning into new opportunities.<br> | | | | | | Mining Test Quantities with Exchangeability: Bayesian Reversibility<br>Entrepreneurial decisions often appear path-dependent, locking ventures into seemingly irreversible directions. This paper introduces three interconnected innovations to address this challenge. We begin with a theoretical foundation - the Symbolic-Algorithmic Decomposition - showing how entrepreneurial learning must flow bidirectionally between abstract strategy and concrete operations. We then develop an algorithmic implementation through a coordinated system of agents: A2E explorers who verify exchangeability in uncertain domains, E2K validators who efficiently convert insights into knowledge, and an A2K synthesis mechanism that maintains detailed balance between them. Most importantly, we show how well-designed test quantities can transform path dependence into a probabilistic learning process, where each experiment contributes to an expanding knowledge base regardless of sequence. Using examples of startup in battery and therapeutics domain, we demonstrate how this approach helps ventures discover fundamental latent variables that generate new strategic options, enabling ”least irreversible” progress even under resource constraints. | | | | [[🗄️product2_EDT]]