Skip to main content
Log in

Value of cognitive diversity in science

  • Published:
Synthese Aims and scope Submit manuscript


When should a scientific community be cognitively diverse? This article presents a model for studying how the heterogeneity of learning heuristics used by scientist agents affects the epistemic efficiency of a scientific community. By extending the epistemic landscapes modeling approach introduced by Weisberg and Muldoon, the article casts light on the micro-mechanisms mediating cognitive diversity, coordination, and problem-solving efficiency. The results suggest that social learning and cognitive diversity produce epistemic benefits only when the epistemic community is faced with problems of sufficient difficulty.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Similar content being viewed by others


  1. Diversity in science obviously matters also beyond strictly epistemological concerns. For example, it can help fight epistemic injustice (Fricker 2007) and promote institutional epistemic virtues, e.g., by reducing prejudice and ethnocentrism (Anderson 1995).

  2. This same aspect of cognitive diversity is also the topic of the subsequent epistemic landscape models introduced by Thoma (2015) and Alexander et al. (2015). Zollman (2010), in contrast, deals with a different aspect of diversity. In the context of theory choice, he examines how (a) the flow of information in social networks and (b) the strength of agents’ prior degrees of belief influence the emergence of consensus in theory choice.

  3. For more general discussions of the use of rugged landscapes in problem-solving research, see, e.g., Page (2008, Ch. 1), and Lazer and Friedman (2007).

  4. For a Python replication of Weisberg and Muldoon’s model, and the source code for the simulations in this article, see the repository at Simulations were conducted with n = 50 for each data point. Error bars (when shown) stand for one standard deviation in sample.

  5. Consequently, the average epistemic significance of the population at a particular time (used in Sect. 2.3) is the change in W over one time step, scaled by a constant.

  6. Weisberg and Muldoon (2009, p. 232), suggest implementing such interaction between agents and landscape as a possible extension of their model. Likewise, Thoma (2015, footnote 7) brings up the idea of a modified model where revisiting patches uncovers further epistemic significance, but she does not develop the idea further.

  7. For applications of similar social-learner-explorer strategies in the literature on cultural inheritance and evolution, see Enquist et al. (2007) and Borenstein et al. (2008).

  8. Simulation experiments were also run with agents who, instead of aiming for maximum significance level, aim to maximize the epistemic work done over a future time period, and yet more sophisticated ones who try to take landscape depletion into account by exponentially discounting for distant payoffs based on \(\lambda \). Such decision rules result in a field of vision delimited by a surface of revolution drawn by a non-linear function. Careful analysis of such situations must be left as a task for future work, but in initial experimentation the differences in the shape of the cone did not lead to qualitative changes in the results.

  9. However, as the value of \(\lambda \) becomes smaller, harvesting epistemic significance from a patch becomes slower, and movement on the landscape becomes relatively less costly. Consequently, the choice of a search heuristic becomes less critical. Real research topics with small lambdas would be ones where changing one’s approach is relatively quick compared to the time it takes to produce results by using a chosen approach.

  10. Thanks to the anonymous referee for insisting on the use of alternative measures of epistemic success.

  11. Due to the nature of the model (see the discussion in Sect. 5), the qualitative results from the model should not depend on the choice of particular point values for the parameters. The scaling of the parameter space was chosen mainly for convenience and for its continuity with the EL model.

  12. Animations of individual simulation runs with different kinds of populations can be found in the online supplementary materials.

  13. As the ruggedness of the landscape increases even more, local search heuristics generally become less and less useful, and even the diverse communities make little progress. Such landscapes can be seen to represent research problems beyond the cognitive capacities of the scientist agents, where attaining significant results becomes increasingly a matter of luck.

  14. As document 1 in the supplementary materials shows, these results hold remarkably well across beta values ranging from 50 to 300, and across different times of measurement.

  15. By herding, I refer to undesirable behavior where agents do what others do even in situations where they should be relying on their own information (Banerjee 1992).

  16. Measuring epistemic progress\(^*\) (see supplementary material, documents 3 and 4) suggests a further advantage of diversity. Unlike populations of social learners, given enough time, diverse populations and populations of individual learners do not leave behind unexamined significant patches on the landscape. This is due to their local search strategy: Agents conducting individual learning only leave a neighborhood of patches once all its patches have been depleted. Hence, in this way, diversity can make the population of scientists more pedantic in its work.

  17. As full results reported in document 4 in the supplementary materials confirm, the ordering in Fig. 5 is stable across the whole range of examined \(\beta \) values. However, for the higher ruggedness values (\(\beta \in [200,300]\)), adding some extra individual learners can lead to a small additional payoff in last stages of the 1000-round simulation run.

  18. The interaction effects between task difficulty and the distribution of social learning thresholds shows in Table 2, where no column dominates the others. For example, at different values of \(\lambda \), pure social learning can be the best, second, or even the worst learning heuristic among the three.


  • Alexander, J. M., Himmelreich, J., & Thompson, C. (2015). Epistemic landscapes, optimal search, and the division of cognitive labor. Philosophy of Science, 82(3), 424–453.

    Article  Google Scholar 

  • Anderson, E. (1995). Knowledge, human Interests, and objectivity in feminist epistemology. Philosophical Topics, 23(2), 27–58.

    Article  Google Scholar 

  • Banerjee, A. V. (1992). A simple model of herd behavior. The Quarterly Journal of Economics, 107(3), 797–817. 05045.

    Article  Google Scholar 

  • Beisbart, C. (2012). How can computer simulations produce new knowledge? European Journal for Philosophy of Science, 2(3), 395–434.

    Article  Google Scholar 

  • Borenstein, E., Feldman, M. W., & Aoki, K. (2008). Evolution of learning in fluctuating environments: When selection favors both social and exploratory individual learning. Evolution, 62(3), 586–602.

    Article  Google Scholar 

  • D’Agostino, F. (2009). From the organization to the division of cognitive labor. Politics, Philosophy & Economics, 8(1), 101–129.

    Article  Google Scholar 

  • De Langhe, R. (2014). A unified model of the division of cognitive labor. Philosophy of Science, 81(3), 444–459.

    Article  Google Scholar 

  • Enquist, M., Eriksson, K., & Ghirlanda, S. (2007). Critical social learning: A solution to Rogers’s paradox of nonadaptive culture. American Anthropologist, 109(4), 727–734.

    Article  Google Scholar 

  • Fricker, M. (2007). Epistemic injustice: Power and the ethics of knowing. Oxford: Oxford University Press.

    Book  Google Scholar 

  • Goldman, A. (2011). A guide to social epistemology. In A. Goldman & D. Whitecomb (Eds.), Social epistemology: Essential readings (pp. 11–37). Oxford: Oxford University Press.

    Google Scholar 

  • Grimm, V., & Berger, U. (2016). Robustness analysis: Deconstructing computational models for ecological theory and applications. Ecological Modelling, 326, 162–167.

    Article  Google Scholar 

  • Hull, D. L. (1988). Science as a process: An evolutionary account of the social and conceptual development of science. Science and its conceptual foundations. Chicago, London: University of Chicago Press.

    Book  Google Scholar 

  • Johnson, S. (2011). Where good ideas come from: The seven patterns of innovation. London: Penguin.

    Google Scholar 

  • Kauffman, S., & Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. Journal of Theoretical Biology, 128(1), 11–45.

    Article  Google Scholar 

  • Kitcher, P. (1990). The division of cognitive labor. Journal of Philosophy, 87(1), 5–22.

    Article  Google Scholar 

  • Kitcher, P. (1993). The advancement of science: Science without legend, objectivity without illusions. New York, Oxford: Oxford University Press.

    Google Scholar 

  • Klarreich, E. (2013). Unheralded mathematician bridges the prime gap. Quanta Magazine

  • Lazer, D., & Friedman, A. (2007). The network structure of exploration and exploitation. Administrative Science Quarterly, 52(4), 667–694.

    Article  Google Scholar 

  • Longino, H. E. (1990). Science as social knowledge: Values and objectivity in scientific inquiry. Princeton: Princeton University Press.

    Google Scholar 

  • Longino, H. E. (2002). The fate of knowledge. Princeton, Oxford: Princeton University Press.

    Google Scholar 

  • MacArthur-Foundation (2014). MacArthur fellows: Yitang Zhang.

  • Marchi, Sd, & Page, S. E. (2014). Agent-based models. Annual Review of Political Science, 17(1), 1–20.

    Article  Google Scholar 

  • Mayo-Wilson, C., Zollman, K. J. S., & Danks, D. (2011). The independence thesis: When individual and social epistemology diverge. Philosophy of Science, 78(4), 653–677.

    Article  Google Scholar 

  • Muldoon, R. (2013). Diversity and the division of cognitive labor. Philosophy Compass, 8(2), 117–125.

    Article  Google Scholar 

  • Novak, M. (2014). Tesla and the lone inventor myth.

  • Page, S. E. (2008). The difference: How the power of diversity creates better groups, firms, schools, and societies (paperback ed.). Princeton, Woodstock: Princeton University Press.

    Google Scholar 

  • Polanyi, M. (1962). The republic of science. Minerva, 1(1), 54–73.

    Article  Google Scholar 

  • Russell, S. J., & Norvig, P. (2003). Artificial intelligence: A modern approach (2nd ed.). Prentice Hall series in artificial intelligence. Upper Saddle River, London: Prentice Hall/Pearson Education.

  • Solomon, M. (2006). Groupthink versus the wisdom of crowds. Southern Journal of Philosophy, 44(Supplement), 28–42.

    Article  Google Scholar 

  • Strevens, M. (2003). The role of the priority rule in science. Journal of Philosophy, 100(2), 55–79.

    Article  Google Scholar 

  • Thoma, J. (2015). The epistemic division of labor revisited. Philosophy of Science, 82(3), 454–472.

    Article  Google Scholar 

  • Weisberg, M. (2010). New approaches to the division of cognitive labor. In J. Busch & P. D. Magnus (Eds.), New waves in philosophy of science. Basingstoke: Palgrave Macmillan.

    Google Scholar 

  • Weisberg, M., & Muldoon, R. (2009). Epistemic landscapes and the division of cognitive labor. Philosophy of Science, 76(2), 225–252.

    Article  Google Scholar 

  • Wright, S. (1932). The roles of mutation, inbreeding, crossbreeding, and selection in evolution. In Proceedings of the Sixth International Congress on Genetics, pp. 355–366

  • Ylikoski, P., & Aydinonat, N. E. (2014). Understanding with theoretical models. Journal of Economic Methodology, 21(1), 19–36.

    Article  Google Scholar 

  • Zollman, K. J. S. (2010). The epistemic benefit of transient diversity. Erkenntnis, 72(1), 17–35.

    Article  Google Scholar 

Download references


I would like to thank the anonymous referees for their useful comments and suggestions on earlier drafts of the paper. I am also thankful to Manuela Fernández Pinto, Marion Godman, Jaakko Kuorikoski, Otto Lappi, Caterina Marchionni, Carlo Martini, and Petri Ylikoski for helpful discussions about the paper and about epistemic landscapes modeling in general. The paper also benefited from comments by the participants at the Agent-based Models in Philosophy conference at LMU Munich, TINT brown bag seminar, and the cognitive science research seminar at University of Helsinki. This research has been financially supported by the Academy of Finland and the University of Helsinki.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Samuli Pöyhönen.

Electronic supplementary material

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Pöyhönen, S. Value of cognitive diversity in science. Synthese 194, 4519–4540 (2017).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: