Published Works
Heshmati, M. and Csaszar, F. A. 2024. “Learning strategic representations: Exploring the effects of taking a strategy course." Organization Science. 35(2) 1383-1399.
[Download paper]
Despite the popularity of strategy courses and the fact that managers make consequential decisions using ideas they learn in such courses, few studies have examined the learning outcomes of taking a strategy course — a research gap most likely due to the methodological challenges of measuring these outcomes in realistic ways. This paper provides a large-sample study of what individuals learn from taking a strategy course and how those learning outcomes depend on individual characteristics. We examine how 2,269 MBA students evaluate real-world video cases before and after taking the MBA core strategy course at a large US business school. We document several changes in their performance, mental representations, and self-perceptions. Among other findings, we show that taking a strategy course improves strategic decision-making, increases the depth of mental representations and the attention paid to broader industry and competitive concerns, and boosts students’ confidence while making them more aware of the uncertainty pervading strategic decisions. We also find that the magnitude and significance of these changes are associated with individual characteristics such as cognitive ability, prior knowledge, and gender.
[Download paper]
Despite the popularity of strategy courses and the fact that managers make consequential decisions using ideas they learn in such courses, few studies have examined the learning outcomes of taking a strategy course — a research gap most likely due to the methodological challenges of measuring these outcomes in realistic ways. This paper provides a large-sample study of what individuals learn from taking a strategy course and how those learning outcomes depend on individual characteristics. We examine how 2,269 MBA students evaluate real-world video cases before and after taking the MBA core strategy course at a large US business school. We document several changes in their performance, mental representations, and self-perceptions. Among other findings, we show that taking a strategy course improves strategic decision-making, increases the depth of mental representations and the attention paid to broader industry and competitive concerns, and boosts students’ confidence while making them more aware of the uncertainty pervading strategic decisions. We also find that the magnitude and significance of these changes are associated with individual characteristics such as cognitive ability, prior knowledge, and gender.
Csaszar, F. A., Rosenkranz, N., and Heshmati, M. 2024. “External representations in strategic decision making: Understanding strategy’s reliance on visuals.” Strategic Management Journal. 45(11) 2191-2226.
[Download paper]
External representations, particularly visuals, are important in strategic decision-making. However, their pervasiveness and impact are not well understood in the strategy literature. Based on cognitive science research, we identify four cognitive functions crucial to strategic decision-making that benefit from using external representations. We also propose a conceptual model and propositions that explain how the quality of strategic decision-making depends on the interactions among task environment, external representations, and managers. We show that external representations influence in predictable ways the boundedly-rational process of searching for new strategies. Key determinants include the manager’s representational capability and the usability and malleability of the external representation. We discuss implications for users, designers, and teachers of external representations in strategy, as well as suggest avenues for future research.
[Download paper]
External representations, particularly visuals, are important in strategic decision-making. However, their pervasiveness and impact are not well understood in the strategy literature. Based on cognitive science research, we identify four cognitive functions crucial to strategic decision-making that benefit from using external representations. We also propose a conceptual model and propositions that explain how the quality of strategic decision-making depends on the interactions among task environment, external representations, and managers. We show that external representations influence in predictable ways the boundedly-rational process of searching for new strategies. Key determinants include the manager’s representational capability and the usability and malleability of the external representation. We discuss implications for users, designers, and teachers of external representations in strategy, as well as suggest avenues for future research.
Working papers
Heshmati, M. and Pahnke, E. “Firms as influencers: Shaping industries through search and collaboration.” [Under review]
Allen, R., Heshmati, M., Lenox, M., McDonald, R., and Perez, M.* "LLMs as belief reinforcers: How human mental representations shape AI-augmented strategy" [Under review]
Large language models (LLMs) excel at many tasks, yet their value for highly uncertain and complex strategy problems is unclear. We study how LLM use affects strategic decision-making, and how humans' mental representations shape these effects. In a 2×2 classroom experiment (N≈200) using the Back Bay Battery simulation, we vary access to an LLM (OpenAI’s o4-mini) and exposure to disruption theory. Among active LLM users, outcomes hinge on their mental representations: without the theory exposure, greater LLM use increases investment in the legacy business; with it, LLM use shifts investment toward the emerging technology. Along with qualitative post-hoc analysis of chat logs, these results show that LLMs act as belief reinforcers, implying that human mental representations are a key contingency in understanding AI-augmented strategic decision-making.
Heshmati, M. and Wang, M. Z.* "Market responses to AI statements in creative domains."
Generative artificial intelligence (AI) can raise entrepreneurial productivity, yet market response also depends on how products are evaluated, particularly in creative domains where human craft is central. Prior work emphasizes AI's internal benefits but offers limited evidence on how market evaluations of products change when entrepreneurs make visible statements about AI use. As platforms increasingly require AI-use disclosures, understanding the market consequences of such positioning becomes critical. We examine this using Kickstarter's AI-disclosure policy, analyzing 42,745 crowdfunding campaigns launched after its introduction. We compare projects that disclose AI use, projects that adopt anti-AI narratives, and projects that remain silent about AI. Relative to silent projects, AI-disclosing campaigns attract 33% fewer backers and are 11.9 percentage points less likely to reach their funding goals, whereas anti-AI narratives are associated with 29% more backers and a 3.9 percentage points higher success rate. Among AI-disclosing projects, framing AI as assisting rather than replacing human work substantially reduces the penalty. The results show that visible statements about AI use serve as consequential elements of market positioning in creative domains, creating tensions between transparency and market performance.
Wang, M. Z. and Heshmati, M. “A founder by any other name: Endowed luck and salience bias in early-stage venture financing.”
Early-stage venture financing unfolds in crowded markets where quality signals are thin, leading investors to rely on founding-team characteristics. We examine how endowed luck, operating through salience bias, shapes fundraising outcomes and how the revealing of quality over time moderates this effect. We derive formal propositions and test them on 155,086 U.S. startups and each founding team’s rarest surname—an unchosen, salient marker uncorrelated with capability. Each order-of-magnitude increase in commonness reduces the odds of raising a first round by 3.4%. As information about quality accumulates, this negative association weakens and even reverses, as low-quality ventures that survived earlier on luck exit later at higher rates. Securing a high-centrality investor early, when salience bias is most acute, benefits unlucky teams more than their luckier peers.
- Runner-up, 2023 Strategic Management Society Conference Behavioral Strategy Interest Group Best Paper Prize
Allen, R., Heshmati, M., Lenox, M., McDonald, R., and Perez, M.* "LLMs as belief reinforcers: How human mental representations shape AI-augmented strategy" [Under review]
Large language models (LLMs) excel at many tasks, yet their value for highly uncertain and complex strategy problems is unclear. We study how LLM use affects strategic decision-making, and how humans' mental representations shape these effects. In a 2×2 classroom experiment (N≈200) using the Back Bay Battery simulation, we vary access to an LLM (OpenAI’s o4-mini) and exposure to disruption theory. Among active LLM users, outcomes hinge on their mental representations: without the theory exposure, greater LLM use increases investment in the legacy business; with it, LLM use shifts investment toward the emerging technology. Along with qualitative post-hoc analysis of chat logs, these results show that LLMs act as belief reinforcers, implying that human mental representations are a key contingency in understanding AI-augmented strategic decision-making.
Heshmati, M. and Wang, M. Z.* "Market responses to AI statements in creative domains."
Generative artificial intelligence (AI) can raise entrepreneurial productivity, yet market response also depends on how products are evaluated, particularly in creative domains where human craft is central. Prior work emphasizes AI's internal benefits but offers limited evidence on how market evaluations of products change when entrepreneurs make visible statements about AI use. As platforms increasingly require AI-use disclosures, understanding the market consequences of such positioning becomes critical. We examine this using Kickstarter's AI-disclosure policy, analyzing 42,745 crowdfunding campaigns launched after its introduction. We compare projects that disclose AI use, projects that adopt anti-AI narratives, and projects that remain silent about AI. Relative to silent projects, AI-disclosing campaigns attract 33% fewer backers and are 11.9 percentage points less likely to reach their funding goals, whereas anti-AI narratives are associated with 29% more backers and a 3.9 percentage points higher success rate. Among AI-disclosing projects, framing AI as assisting rather than replacing human work substantially reduces the penalty. The results show that visible statements about AI use serve as consequential elements of market positioning in creative domains, creating tensions between transparency and market performance.
Wang, M. Z. and Heshmati, M. “A founder by any other name: Endowed luck and salience bias in early-stage venture financing.”
Early-stage venture financing unfolds in crowded markets where quality signals are thin, leading investors to rely on founding-team characteristics. We examine how endowed luck, operating through salience bias, shapes fundraising outcomes and how the revealing of quality over time moderates this effect. We derive formal propositions and test them on 155,086 U.S. startups and each founding team’s rarest surname—an unchosen, salient marker uncorrelated with capability. Each order-of-magnitude increase in commonness reduces the odds of raising a first round by 3.4%. As information about quality accumulates, this negative association weakens and even reverses, as low-quality ventures that survived earlier on luck exit later at higher rates. Securing a high-centrality investor early, when salience bias is most acute, benefits unlucky teams more than their luckier peers.
Early stage works
Li., C. and Heshmati, M.* “Borrowed cognition: How AI exposure impacts idea generation and evaluation”
Generative AI tools are increasingly embedded in organizations' innovation processes. Although such tools--particularly large language models (LLMs)--extend cognition during use, we know little about how the “naked” mind performs once the tool is removed. We ask how exposure to AI during idea generation alters individuals’ subsequent unaided ability to generate and evaluate ideas, and how an individual’s division of labor with the tool conditions these effects. We run an experiment in which participants are randomly assigned LLM access for an initial idea-generation task; the tool is then removed before a second idea-generation task in a new domain and a subsequent evaluation task. Results show that greater delegation to an LLM during brainstorming is associated with declines in later generation quality and evaluation quality (via increased omission errors), whereas lenient screening of LLM output is associated with a larger decline in later generation quality but only marginal effects on evaluation quality. Thus, both forms of reliance on AI degrade post-AI generation quality, while degradation in evaluation skill is concentrated in dependency during brainstorming. We develop a framework explaining how styles of delegation between humans and LLMs shape decision performance and outline implications and a research agenda for cognition, AI, and organizational design. We contribute to strategic decision-making research by showing that brief AI exposure can have lingering cognitive effects and by identifying design levers to help preserve the “naked mind.”
Heshmati, M. and Sirmon, D. "Search and decision making flexibility in transient competitive contexts: Evidence from the NHL"
Drawing on the Carnegie School tradition, this study examines how search concentration (a firm’s focus on narrow sets of alternatives) and decision-making flexibility (the ability to rapidly adjust to evolving circumstances) jointly influence performance in dynamic contexts. Analyzing National Hockey League play-by-play data, we find that both factors independently boost performance, though their interaction varies with context. Under transient advantage—a brief window of opportunity—concentrated search and flexibility amplify each other’s benefits, allowing firms to exploit the window while remaining agile. However, when facing a transient disadvantage, their combination increases vulnerability to rival attacks, suggesting that reactive adjustments in a narrow search may trigger overcorrection. Our findings emphasize aligning search focus with decision-making style and demonstrate the context-dependent nature of search in shaping success.
Heshmati, M. and Choe, S. “Hiding Their True Colors: Exploring the Interplay of Stigmatized Identity and Competitive Positioning”
* Denotes equal coauthorship.
Li., C. and Heshmati, M.* “Borrowed cognition: How AI exposure impacts idea generation and evaluation”
Generative AI tools are increasingly embedded in organizations' innovation processes. Although such tools--particularly large language models (LLMs)--extend cognition during use, we know little about how the “naked” mind performs once the tool is removed. We ask how exposure to AI during idea generation alters individuals’ subsequent unaided ability to generate and evaluate ideas, and how an individual’s division of labor with the tool conditions these effects. We run an experiment in which participants are randomly assigned LLM access for an initial idea-generation task; the tool is then removed before a second idea-generation task in a new domain and a subsequent evaluation task. Results show that greater delegation to an LLM during brainstorming is associated with declines in later generation quality and evaluation quality (via increased omission errors), whereas lenient screening of LLM output is associated with a larger decline in later generation quality but only marginal effects on evaluation quality. Thus, both forms of reliance on AI degrade post-AI generation quality, while degradation in evaluation skill is concentrated in dependency during brainstorming. We develop a framework explaining how styles of delegation between humans and LLMs shape decision performance and outline implications and a research agenda for cognition, AI, and organizational design. We contribute to strategic decision-making research by showing that brief AI exposure can have lingering cognitive effects and by identifying design levers to help preserve the “naked mind.”
Heshmati, M. and Sirmon, D. "Search and decision making flexibility in transient competitive contexts: Evidence from the NHL"
Drawing on the Carnegie School tradition, this study examines how search concentration (a firm’s focus on narrow sets of alternatives) and decision-making flexibility (the ability to rapidly adjust to evolving circumstances) jointly influence performance in dynamic contexts. Analyzing National Hockey League play-by-play data, we find that both factors independently boost performance, though their interaction varies with context. Under transient advantage—a brief window of opportunity—concentrated search and flexibility amplify each other’s benefits, allowing firms to exploit the window while remaining agile. However, when facing a transient disadvantage, their combination increases vulnerability to rival attacks, suggesting that reactive adjustments in a narrow search may trigger overcorrection. Our findings emphasize aligning search focus with decision-making style and demonstrate the context-dependent nature of search in shaping success.
Heshmati, M. and Choe, S. “Hiding Their True Colors: Exploring the Interplay of Stigmatized Identity and Competitive Positioning”
- Research sponsored by the UW Consulting and Business Development Center (CBDC)
* Denotes equal coauthorship.