We revisit the competency trap and reexamine when it occurs. We show that a bias against alternat... more We revisit the competency trap and reexamine when it occurs. We show that a bias against alternatives that improve with practice does not require that learning is myopic in the sense of lacking foresight or failing to explore. The same bias occurs even if learners engage in substantial exploration and have foresight. In fact, we demonstrate that even a rational and foresighted learner, who follow an optimal strategy for balancing exploration and exploitation, will learn to prefer alternatives with initially high payoffs that decrease with practice over alternatives, with identical expected values, that have initially low payoffs that increase with practice. Our results show that a bias against alternatives that improve with practice is due to an asymmetry in error correction rather than to myopic learning. The implication is that a wide range of selection systems, even optimally designed ones, will be biased against late-bloomers.
People frequently consult average ratings on online recommendation platforms before making consum... more People frequently consult average ratings on online recommendation platforms before making consumption decisions. Research on the wisdom-of-the-crowd phenomenon suggests that average ratings provide unbiased quality estimates. Yet we argue that the process by which average ratings are updated creates a systematic bias. In analyses of more than 80 million online ratings, we found that items with high average ratings tend to attract more additional ratings than items with low average ratings. We call this asymmetry in how average ratings are updated endogenous crowd formation. Using computer simulations, we showed that it implies the emergence of a negative bias in average ratings. This bias affects items with few ratings particularly strongly, which leads to ranking mistakes. The average-rating rankings of items with few ratings are worse than their quality rankings. We found evidence for the predicted pattern of biases in an experiment and in analyses of large online-rating data sets.
A key function of categories is to help predictions about unobserved features of objects. At the ... more A key function of categories is to help predictions about unobserved features of objects. At the same time, humans are often in situations where the categories of the objects they perceive are uncertain. In an influential paper, Anderson (Psychological Review, 98(3), 409-429, 1991) proposed a rational model for feature inferences with uncertain categorization. A crucial feature of this model is the conditional independence assumption-it assumes that the within category feature correlation is zero. In prior research, this model has been found to provide a poor fit to participants' inferences. This evidence is restricted to task environments inconsistent with the conditional independence assumption. Currently available evidence thus provides little information about how this model would fit participants' inferences in a setting with conditional independence. In four experiments based on a novel paradigm and one experiment based on an existing paradigm, we assess the performance of Anderson's model under conditional independence. We find that this model predicts participants' inferences better than competing models. One model assumes that inferences are based on just the most likely Electronic supplementary material The online version of this article (
We demonstrate that a sampling-based mechanism can offer an alternative explanation for belief sy... more We demonstrate that a sampling-based mechanism can offer an alternative explanation for belief synchronization in social groups and the persistence of collective illusions. Our model assumes that people are more likely to sample popular alternatives than unpopular alternatives. We show that this mechanism is sufficient to explain belief synchronization: a strong majority of opinions will likely emerge in favor of one alternative. The reason is that the group is unlikely to move away from a state in which one alternative is very unpopular. If by chance most people come to dislike alternative A, they are all unlikely to sample it again and their opinions of A remain the same. When A is in fact the best alternative, a collective illusion has emerged because people mistakenly believe that a suboptimal alternative is the best. Our model implies that such a collective illusion is persistent. The model thus offers an existence proof that a collective illusion can occur even in settings where people do not infer that popular alternatives are better. The model also casts new light on the effect of online recommendation systems on attitude homogenization and the effect of majority voting on beliefs and attitudes.
Societies sometimes stick to the status quo instead of switching to superior technologies and ins... more Societies sometimes stick to the status quo instead of switching to superior technologies and institutions. Existing explanations often attribute this to a coordination failure due to payoff externalities: people may know that another alternative is superior but nobody has an incentive to switch unless many others do so. We show that a simple learning argument can provide an alternative explanation. When people learn about the alternatives from their own experiences but tend to adopt the behaviors of others, they will mistakenly learn to believe that a popular alternative is superior to a better, but unpopular alternative. Our model neither assumes that agents engage in motivated cognition nor that they transmit mistaken information to others. Rather, it emphasizes the role of a fundamental asymmetry in access to information about popular versus unpopular alternatives. Our model thus provides a novel, sampling-based, explanation of how conformity in behavior can lead to private acce...
Negro, and Anand Swaminathan. We thank Raymond Freeman, Jr. for copies of the Freeman Reports, an... more Negro, and Anand Swaminathan. We thank Raymond Freeman, Jr. for copies of the Freeman Reports, and Robert Abraham, who continued the business upon Freeman's retirement, for his help in understanding the industry. We also thank the Charles Babbage Institute (especially Carrie Seib) and the Hagley Museum and Library (especially Carol Lockman and Marjorie McNinch) for access to their respective collections. Byung-Soo Kim worked as a research assistant in the early stages of this project.
Bentley et al. make the deliberate choice to blur the distinction between learning and decision m... more Bentley et al. make the deliberate choice to blur the distinction between learning and decision making. This obscures the social influence mechanisms that operate in the various empirical settings that their map aims to categorize. Useful policy prescriptions, however, require an accurate understanding of the social influence mechanisms that underlie the dynamics of popularity.
New products, services, and ideas are often evaluated more favorably than similar but older ones.... more New products, services, and ideas are often evaluated more favorably than similar but older ones. Although several explanations of this phenomenon have been proposed, we identify an overlooked asymmetry in information about new and old items that emerges when people seek positive experiences and learn about the qualities of (noisy) alternatives by experiencing them. The reason for the asymmetry is that people avoid rechoosing alternatives that previously led to poor outcomes; hence, additional feedback on their qualities is precluded. Negative quality estimates, even when caused by noise, thus tend to persist. This negative bias takes time to develop, and affects old alternatives more strongly than similar but newer alternatives. We analyze a simple learning model and demonstrate the process by which people would tend to evaluate a new alternative more positively than an older alternative with the same payoff distribution. The results from two experimental studies ( Ns = 769 and 805...
Ranking algorithms play a crucial role in online platforms ranging from search engines to recomme... more Ranking algorithms play a crucial role in online platforms ranging from search engines to recommender systems. In this paper, we identify a surprising consequence of popularity-based rankings: the fewer the items reporting a given signal, the higher the share of the overall traffic they collectively attract. This few-get-richer effect emerges in settings where there are few distinct classes of items (e.g., left-leaning news sources versus right-leaning news sources), and items are ranked based on their popularity. We demonstrate analytically that the few-get-richer effect emerges when people tend to click on top-ranked items and have heterogeneous preferences for the classes of items. Using simulations, we analyze how the strength of the effect changes with assumptions about the setting and human behavior. We also test our predictions experimentally in an online experiment with human participants. Our findings have important implications to understand the spread of misinformation.
We demonstrate a paradox of selection: the average level of skill among the survivors of selectio... more We demonstrate a paradox of selection: the average level of skill among the survivors of selection may initially increase but eventually decrease. This result occurs in a simple model in which performance is not frequency dependent, there are no delayed effects, and skill is unrelated to risk-taking. The performance of an agent in any given period equals a skill component plus a noise term. We show that the average skill of survivors eventually decreases when the noise terms in consecutive periods are dependent and drawn from a distribution with a "long" tail-a sub-class of heavy-tailed distributions. This result occurs because only agents with extremely high level of performance survive many periods, and extreme performance is not diagnostic of high skill when the noise term is drawn from a long-tailed distribution. History: This paper has been accepted for the Strategy Science Special Issue on Evolutionary Perspectives on Strategy.
Empirical evidence about the relation between organizational age and failure is mixed, and theore... more Empirical evidence about the relation between organizational age and failure is mixed, and theoretical explanations are conflicting. We show that a simple model of organizational evolution can explain the main patterns of age dependence and reconcile the apparently conflicting theoretical predictions. In our framework, the predicted pattern of age dependence depends crucially on the quality of organizational performance immediately after founding and its subsequent evolution, which in turn depends on the intensity of competition. In developing our theory, we clarify issues of levels of analysis as well as the relations between organizational fitness, endowment, organizational capital, and the hazard of failure. We show that once organizational learning is considered, founding conditions affect the fate of organizations in ways more complex than previously acknowledged. We illustrate how the predictions of our theory can be tested empirically and evaluate the effect of aging on the m...
We revisit the competency trap and reexamine when it occurs. We show that a bias against alternat... more We revisit the competency trap and reexamine when it occurs. We show that a bias against alternatives that improve with practice does not require that learning is myopic in the sense of lacking foresight or failing to explore. The same bias occurs even if learners engage in substantial exploration and have foresight. In fact, we demonstrate that even a rational and foresighted learner, who follow an optimal strategy for balancing exploration and exploitation, will learn to prefer alternatives with initially high payoffs that decrease with practice over alternatives, with identical expected values, that have initially low payoffs that increase with practice. Our results show that a bias against alternatives that improve with practice is due to an asymmetry in error correction rather than to myopic learning. The implication is that a wide range of selection systems, even optimally designed ones, will be biased against late-bloomers.
People frequently consult average ratings on online recommendation platforms before making consum... more People frequently consult average ratings on online recommendation platforms before making consumption decisions. Research on the wisdom-of-the-crowd phenomenon suggests that average ratings provide unbiased quality estimates. Yet we argue that the process by which average ratings are updated creates a systematic bias. In analyses of more than 80 million online ratings, we found that items with high average ratings tend to attract more additional ratings than items with low average ratings. We call this asymmetry in how average ratings are updated endogenous crowd formation. Using computer simulations, we showed that it implies the emergence of a negative bias in average ratings. This bias affects items with few ratings particularly strongly, which leads to ranking mistakes. The average-rating rankings of items with few ratings are worse than their quality rankings. We found evidence for the predicted pattern of biases in an experiment and in analyses of large online-rating data sets.
A key function of categories is to help predictions about unobserved features of objects. At the ... more A key function of categories is to help predictions about unobserved features of objects. At the same time, humans are often in situations where the categories of the objects they perceive are uncertain. In an influential paper, Anderson (Psychological Review, 98(3), 409-429, 1991) proposed a rational model for feature inferences with uncertain categorization. A crucial feature of this model is the conditional independence assumption-it assumes that the within category feature correlation is zero. In prior research, this model has been found to provide a poor fit to participants' inferences. This evidence is restricted to task environments inconsistent with the conditional independence assumption. Currently available evidence thus provides little information about how this model would fit participants' inferences in a setting with conditional independence. In four experiments based on a novel paradigm and one experiment based on an existing paradigm, we assess the performance of Anderson's model under conditional independence. We find that this model predicts participants' inferences better than competing models. One model assumes that inferences are based on just the most likely Electronic supplementary material The online version of this article (
We demonstrate that a sampling-based mechanism can offer an alternative explanation for belief sy... more We demonstrate that a sampling-based mechanism can offer an alternative explanation for belief synchronization in social groups and the persistence of collective illusions. Our model assumes that people are more likely to sample popular alternatives than unpopular alternatives. We show that this mechanism is sufficient to explain belief synchronization: a strong majority of opinions will likely emerge in favor of one alternative. The reason is that the group is unlikely to move away from a state in which one alternative is very unpopular. If by chance most people come to dislike alternative A, they are all unlikely to sample it again and their opinions of A remain the same. When A is in fact the best alternative, a collective illusion has emerged because people mistakenly believe that a suboptimal alternative is the best. Our model implies that such a collective illusion is persistent. The model thus offers an existence proof that a collective illusion can occur even in settings where people do not infer that popular alternatives are better. The model also casts new light on the effect of online recommendation systems on attitude homogenization and the effect of majority voting on beliefs and attitudes.
Societies sometimes stick to the status quo instead of switching to superior technologies and ins... more Societies sometimes stick to the status quo instead of switching to superior technologies and institutions. Existing explanations often attribute this to a coordination failure due to payoff externalities: people may know that another alternative is superior but nobody has an incentive to switch unless many others do so. We show that a simple learning argument can provide an alternative explanation. When people learn about the alternatives from their own experiences but tend to adopt the behaviors of others, they will mistakenly learn to believe that a popular alternative is superior to a better, but unpopular alternative. Our model neither assumes that agents engage in motivated cognition nor that they transmit mistaken information to others. Rather, it emphasizes the role of a fundamental asymmetry in access to information about popular versus unpopular alternatives. Our model thus provides a novel, sampling-based, explanation of how conformity in behavior can lead to private acce...
Negro, and Anand Swaminathan. We thank Raymond Freeman, Jr. for copies of the Freeman Reports, an... more Negro, and Anand Swaminathan. We thank Raymond Freeman, Jr. for copies of the Freeman Reports, and Robert Abraham, who continued the business upon Freeman's retirement, for his help in understanding the industry. We also thank the Charles Babbage Institute (especially Carrie Seib) and the Hagley Museum and Library (especially Carol Lockman and Marjorie McNinch) for access to their respective collections. Byung-Soo Kim worked as a research assistant in the early stages of this project.
Bentley et al. make the deliberate choice to blur the distinction between learning and decision m... more Bentley et al. make the deliberate choice to blur the distinction between learning and decision making. This obscures the social influence mechanisms that operate in the various empirical settings that their map aims to categorize. Useful policy prescriptions, however, require an accurate understanding of the social influence mechanisms that underlie the dynamics of popularity.
New products, services, and ideas are often evaluated more favorably than similar but older ones.... more New products, services, and ideas are often evaluated more favorably than similar but older ones. Although several explanations of this phenomenon have been proposed, we identify an overlooked asymmetry in information about new and old items that emerges when people seek positive experiences and learn about the qualities of (noisy) alternatives by experiencing them. The reason for the asymmetry is that people avoid rechoosing alternatives that previously led to poor outcomes; hence, additional feedback on their qualities is precluded. Negative quality estimates, even when caused by noise, thus tend to persist. This negative bias takes time to develop, and affects old alternatives more strongly than similar but newer alternatives. We analyze a simple learning model and demonstrate the process by which people would tend to evaluate a new alternative more positively than an older alternative with the same payoff distribution. The results from two experimental studies ( Ns = 769 and 805...
Ranking algorithms play a crucial role in online platforms ranging from search engines to recomme... more Ranking algorithms play a crucial role in online platforms ranging from search engines to recommender systems. In this paper, we identify a surprising consequence of popularity-based rankings: the fewer the items reporting a given signal, the higher the share of the overall traffic they collectively attract. This few-get-richer effect emerges in settings where there are few distinct classes of items (e.g., left-leaning news sources versus right-leaning news sources), and items are ranked based on their popularity. We demonstrate analytically that the few-get-richer effect emerges when people tend to click on top-ranked items and have heterogeneous preferences for the classes of items. Using simulations, we analyze how the strength of the effect changes with assumptions about the setting and human behavior. We also test our predictions experimentally in an online experiment with human participants. Our findings have important implications to understand the spread of misinformation.
We demonstrate a paradox of selection: the average level of skill among the survivors of selectio... more We demonstrate a paradox of selection: the average level of skill among the survivors of selection may initially increase but eventually decrease. This result occurs in a simple model in which performance is not frequency dependent, there are no delayed effects, and skill is unrelated to risk-taking. The performance of an agent in any given period equals a skill component plus a noise term. We show that the average skill of survivors eventually decreases when the noise terms in consecutive periods are dependent and drawn from a distribution with a "long" tail-a sub-class of heavy-tailed distributions. This result occurs because only agents with extremely high level of performance survive many periods, and extreme performance is not diagnostic of high skill when the noise term is drawn from a long-tailed distribution. History: This paper has been accepted for the Strategy Science Special Issue on Evolutionary Perspectives on Strategy.
Empirical evidence about the relation between organizational age and failure is mixed, and theore... more Empirical evidence about the relation between organizational age and failure is mixed, and theoretical explanations are conflicting. We show that a simple model of organizational evolution can explain the main patterns of age dependence and reconcile the apparently conflicting theoretical predictions. In our framework, the predicted pattern of age dependence depends crucially on the quality of organizational performance immediately after founding and its subsequent evolution, which in turn depends on the intensity of competition. In developing our theory, we clarify issues of levels of analysis as well as the relations between organizational fitness, endowment, organizational capital, and the hazard of failure. We show that once organizational learning is considered, founding conditions affect the fate of organizations in ways more complex than previously acknowledged. We illustrate how the predictions of our theory can be tested empirically and evaluate the effect of aging on the m...
Uploads
Papers by Gaël Le Mens