Exploration and exploitation in complex search tasks: How feedback influences whether and where human agents search

Abstract

Research Abstract Exploration and exploitation in strategic decision‐making entails decisions about whether and where to search for new alternatives to improve the status quo. Prior research has not explored how decisions about whether to continue search (vs. stop search or satisfice) and where to search (near vs. far) are interrelated. We report laboratory experiment results on how individuals decide whether and where to search in a complex, combinatorial task. We find that different feedback variables influence the decision to stop search from decisions regarding how broadly to search. Our results suggest that not accounting for the decision to continue (or stop) searching, separately from breadth of search, can lead to incorrect predictions regarding how feedback influences search behavior. Managerial Abstract Managers concerned about the performance of their company face two challenges—they have to find out what potential performance is feasible given their business environment and which organizational policies to implement to realize it. We use a stylized laboratory experiment to better understand how feedback from experimentation informs such a learning process. We find that early‐stage feedback has a lasting impact on performance aspirations and managerial expectations about feasible performance. Superior early feedback thereby motivates more sustained experimentation with organizational policies. In contrast, more recent feedback guides the extent to which managers are willing to engage in more radical policy adjustments, especially in the latter stages of a learning process.

The origins of time compression diseconomies

“…Some of these VIF values are above the acceptable value of 10 (Gujarati, 2003, p. 362). To remedy this issue, we adopted the orthogonal polynomial technique (Kleinbaum, Kupper, Muller, & Nizam, 2007; Sribney, 1995; for a recent application, see Billinger, Srikanth, Stieglitz, & Schumacher, 2021). Using this technique, we orthogonalized all regressors with a VIF >10 in the baseline model (Model 2 in Table 3).…”

Who deviates? Technological opportunities, career concern, and inventor's distant search

“…Per "problemistic search" (Cyert & March, 1963;Posen et al, 2018), search is triggered when she faces a problem, and her choice of local or distant search depends on which is more likely to uncover a satisfactory solution. Absent perfect information on global peak, or information or vicarious learning about possible solutions, she relies on feedback from prior search, especially recent ones (Laureiro-Martínez, Brusoni, Canessa, & Zollo, 2015), to form an aspiration level or criterion for deciding what is a satisfactory solution (Billinger et al, 2021;Hu, Blettner, & Bettis, 2011). 2 To substantiate this example, we examined Sony's patenting behavior in the late 1980s and 1990s.…”

“…As Simon (1991, p 125) describes “All learning takes place inside individual human heads; an organization learns … by the learning of its members.” As the inventor searches, her bounded rationality prevents her from knowing where the “globally optimal” solution is (Ahuja et al, 2008; Siggelkow, 2002b), with the severity of this bound depending on her problem‐solving abilities and the problem's structure and complexity (Baumann, Schmidt, & Stieglitz, 2019). Being resource‐constrained in her search (Rivkin, 2000), she follows a “satisficing” rule, capturing solutions as long as they satisfy certain preset criteria or aspiration level (Billinger et al, 2021; Nelson & Winter, 1982).…”

“…This characterization helps conceptually separate local from distant search (Ahuja & Katila, 2004; Billinger et al, 2021). Local search stays within neighborhoods on the landscape nearby the inventor's current solution.…”

“…An inventor's search for new technologies plays a crucial role in building a firm's competitive advantage (Almeida & Kogut, 1999; Toh & Polidoro, 2013). While search has largely been studied at the firm‐level (Ahuja & Katila, 2004; Eggers & Kaul, 2018), recent researchers are calling for a return to viewing search at the individual‐level (Billinger, Srikanth, Stieglitz, & Schumacher, 2021; Lee & Meyer‐Doyle, 2017) so as to uncover more fine‐grained insights on inventors' search behavior. In this view, per the behavioral theory tradition (Cyert & March, 1963; Levinthal, 1997), the inventor is often seen as searching across a rugged landscape in a boundedly rational way, not knowing beforehand where the globally optimal technological solution resides (Ganco & Hoetker, 2009; Siggelkow, 2002b) and deciding from a given point if she should search locally or distantly (Ahuja, Lampert, & Tandon, 2008; Ethiraj & Levinthal, 2004).…”