This is a post by Charlotte Zemmel (Cambridge), a speaker from our Spring 2021 Seminar Series. If you would like to attend a seminar, please sign up using this link. You can also download the full schedule here. You can watch the recording here. Featured image by Daniel Öberg on Unsplash.
We often justify observational claims by appealing to different kinds of evidence collected from a wide range of sources. This is captured in the concept of ‘robustness’; an observation is robust if many different experimental instruments corroborate the same observation in a range of locations.1 The idea at the heart of robustness – that diverse forms of evidence together are stronger than isolated evidence claims – can be extended into a social epistemic argument. A community comprised of diverse epistemic perspectives can produce stronger, more robust claims than a collection of homogenous thinkers. Philosophers term this the ‘epistemic benefit of cognitive diversity’.2
Although this argument is intuitively sound, any effort to purport the specific benefit of cognitive diversity causes one to confront a series of deeper tensions in social epistemology. For example, here, I discuss the partition between those who hold that the best epistemic practices should culminate in a consensus across all agents (the ‘consensors’) and those who hold that consensus shouldn’t be (and often isn’t) the goal of knowledge acquisition (the ‘normative pluralists’). We can reconcile these two diverging philosophical positions in the context of the epistemic benefit of diversity, so long as we recognise two key points. Firstly, the consensors and normative pluralists differ in their propositions for what kind of belief-diversity should be promoted. Specifically, I argue that consensors are concerned with ‘first-order beliefs,’ whereas normative pluralists wish to promote a diversity of ‘second-order’ beliefs. Secondly, arguments for the benefit of diversity must always specify the kinds of epistemic activities to which they refer. By recognising this second argument, I suggest that philosophical defenses of diversity’s epistemic benefit can become ‘sharp.’ By this, I mean that we can tailor arguments on the epistemic advantage of diversity to particular situations by engaging closely with epistemic practices themselves.
Introducing the consensors and the normative pluralists:
Many arguments expressing diversity’s epistemic benefit imply that diversity is necessary to justify a consensus within a community. For example, formal social epistemologists Kevin Zollman3 and Philip Kitcher are concerned with preventing a community from converging on a theory prematurely before more successful theories have been adequately explored. Kitcher suggested that an effective division of labor is necessary when justifying a consensus so that all alternative theories are adequately explored before the consensus is reached.4 Diversity supplies the necessary cognitive differentiation (and workforce) to explore each alternative. In this manner, diversity’s epistemic benefit is closely connected with establishing a justified consensus. Similarly, Boaz Miller considers social diversity necessary when forming a consensus.5 He argues that diversity ensures a consensus is knowledge-based, not the result of collective bias.6
From this brief sketch of the consensor’s arguments, three key benefits of diversity can be identified: diversity generates an effective division of labor, diversity presents alternative theories to the epistemic community, and diversity can help mitigate the negative impacts of bias. However, viewing these arguments from the perspective of consensus-building tells only part of the story of how and why cognitive diversity can be beneficial. In particular, normative pluralists have much to say about why diversity should be fostered that does not rely on any consensus-forming premise. John Stuart Mill’s “tyranny of the majority” can be viewed as a case in point.7 Mill argued that dissensus is always necessary for a democratic society since, without dissenting views, an intellectual homogeneity will always amount to dogmatism.8 Here, Mill promotes a specific kind of justificatory practice; one can only justifiably hold a particular belief if she has compared it with alternatives. Crucially, those alternatives must not be abandoned following her encounter with them, but they too must be preserved: without the constant presence of other options, her belief will no longer be justified.9 Therefore, the benefit of diversity in Mill’s view is to deliver alternatives and prevent bias through dogmatism, but he has interpreted the broader epistemic use of these two functions in contrasting ways to the consensors I explored above. For normative pluralists such as Mill, alternatives are not necessary to justify a collective agreement on what is True or Right, but rather to serve as constant reminders that there are a multiplicity of beliefs that cone can hold at any point in time.
Other normative pluralists oppose the goal of consensus on ontological grounds. For example, Sandra Mitchell argues that we should proceed in science through ‘integrative pluralism’.10 In this method, a range of idealising theories and models are integrated to give a complete causal account of a specific phenomenon. Integrative pluralism’s primary motivation is the inherent limitation on all of our theories, models, and methods, as the real world is much more complex than a single view can depict. If science is to produce successful explanations and predictions, it mustn’t consider a plurality of theories in a given domain to be obstructive to progress. Nor does plurality suggest that we can further reduce each theory into a single, master theory. Instead, Mitchell argues that maintaining pluralism in science is necessary to accurately represent real-world phenomena.
In short, for the normative pluralists, the fact of pluralism suggests that consensus is bad practice. For the consensors, the existence of pluralism signifies a premature subject area. These differing philosophical credences make claims such as, ‘diversity is beneficial for presenting alternatives’ deficient for encouraging diversity-promoting practices. One’s views on the epistemic benefit of offering alternatives depend on where one stands concerning consensus-building, a topic that itself is fraught with philosophical debate.
Interpreting the tension: first-order and second-order beliefs
How are we to move forward in understanding the epistemic benefit of diversity, informed by the more profound tension between the consensors and normative pluralists? Each view has its merits and its limitations. Can we establish an account of beneficial diversity that includes both perspectives? I believe that we can if we recognise that each party is concerned with the diversity of different kinds of beliefs. To illustrate this point, I introduce the concept of first-order and second-order beliefs and their associated justification.11
First-order beliefs are attitudes about the state of the world. For example, “it will rain tomorrow” is a first-order belief. Second-order beliefs are what we evoke to justify our first-order beliefs. “I believe it will rain tomorrow because I checked BBC Weather and BBC Weather is reliable” is a second-order belief. A second-order belief is constructed within a particular agent’s “epistemic system’- “Epistemic systems are social processes generating judgments of truth and falsity.”12 In other words, what one considers to be a viable second-order belief will depend on which social group she belongs.
One can appreciate the epistemic significance of first and second-order beliefs through the context of epistemic disagreement. Suppose two people disagree on the first-order claim that it will rain tomorrow: i.e., a dissenter says, “it is not going to rain tomorrow.” To decide who is right, each enlists all of the second-order beliefs they have to justify their first-order belief: the dissenter responds that “The BBC is unreliable.” Often in cases like these, it is rational to suppose that someone is wrong – either the BBC is reliable or is not. Therefore either it is justified to believe that it will rain tomorrow based on BBC Weather, or it is not. By going through each second-order beliefs, presumably, both will identify where one of them had made an error:
“The BBC is reliable because they use sophisticated weather models.”
“A recent study has shown that these models are inaccurate over 50% of the time.”
“That study was recently challenged by another, which used a larger database.”
A first-order belief’s status is debated through this process by evoking a series of second-order beliefs. However, the notion that second-order beliefs gain their significance from within an epistemic system leaves room for so-called ‘faultless disagreements’,13 where this iterative process of debate breaks down. For two people to debate in the manner introduced above, they need to have a prior agreement on the kinds of things that contribute to the justification of first-order beliefs, e.g., weather models, scientific studies, and large databases. If this meta-consensus of second-order beliefs does not exist, two agents can be entirely justified in holding opposing first-order beliefs. Suppose a third agent comes along and justifies their first-order belief that it will rain tomorrow by referring to the fact that she said a prayer. If her second-order beliefs are formed out of an epistemic system that awards causal properties to prayer, then one clearly would not be able to refer to a weather model to convince her that she is misguided. These two individuals do not disagree in the same way as the two interlocutors above, who were operating in the same epistemic system.
I propose that the consensors exclusively consider the consequences of a diversity of first-order beliefs within well-defined epistemic communities, while the normative pluralists encourage diversity of second-order beliefs. I consider Miriam Solomon’s objection to Mill’s democratic ideal of constant debate between alternatives to justify this distinction. Solomon14 proposes that diverse views are most epistemically valuable when aggregated rather than engaged in critical discourse. Her hesitancy towards Millean pluralism derives from her concern over the epistemic phenomenon of ‘groupthink’. When a collection of diverse agents are urged to engage critically, they subject themselves to socially-driven selection pressures that can cause the abandonment of valuable insights. For example, if a particularly well-known agent is part of the group, her perspectives will be awarded greater epistemic authority, often at the expense of other, lesser-known agents. The community then begins to suffer from ‘groupthink’; a consensus is reached on a particular belief, but not necessarily because that view is best. Conversely, Solomon argues, aggregating each agent’s view, voting-booth-style, allows the community to extract the full benefits of a diverse membership.
While Solomon’s example of groupthink offers a convincing reason to be cautious when proclaiming a given consensus as knowledge-based, she has failed to consider that critical discourse can perform various epistemic functions. Agents can only anticipate critical discourse to lead to an agreement (whether due to groupthink or otherwise) if they expect to be convinced by their peers. For this to happen, agents must interact with individuals who share their second-order beliefs (see example above). Solomon may well be correct in such cases, and it is best to aggregate diverse views from within an epistemic system rather than encourage critical discourse. However, if our interlocutors do not possess a shared standard of second-order beliefs, then critically engaging will surely not lead one party to be convinced into consensus. Furthermore, in such a case, aggregation makes little sense too – on what basis can two beliefs, justified by different second-order claims, be aggregated? Therefore, Solomon’s hesitancy towards critical discourse belongs only to cases where there is a consensus of second-order beliefs and a range of first-order consequences. On the other hand, we can save Mill’s view from the risk of groupthink by interpreting his account as urging critical discourse between agents with differing second-order beliefs, who debate not to be convinced by the other’s views, but to reformulate their own stance, and thereby award themselves with renewed justification.
The result of partitioning the debate along first-order and second-order beliefs is that both the consensors and the normative pluralists must now approach two distinct challenges to the thesis that a diversity of beliefs is an epistemic good. The consensors must explain why it is the case that cognitive diversity is valuable even within a well-defined epistemic system. Conversely, the normative pluralists must explain why it should be the case that diversity of epistemic systems is epistemically beneficial, given that two agents operating within different systems run the risk of falling into faultless disagreements. To this end, I conclude by turning to the importance of contextualising arguments regarding the benefit of diversity.
Epistemic benefit in context
Lord Steyn once said, “In law, context is everything”.15 The same is true for normative arguments on what makes good epistemic practice. Therefore, considering the division of first-order and second-order beliefs is only the first step towards a fully operationalizable account of cognitive diversity’s epistemic benefit. The next step is to consider the consequences of each kind of belief diversity in context. For example, what role can either form of diversity play in situations where an epistemic system is still under construction, such as in so-called exploratory experiments?16 How can second-order belief diversity be beneficial in the face of severe problems of incommensurability between agents operating within different systems?17 Often, we wish to exclude specific belief systems on social-justice grounds. Does Millean critical discourse leave room for blocking potentially harmful epistemic beliefs? The virtue of the consensus’ view is that we save ourselves from the deadlock of faultless disagreements. The vice is that urging a consensus can lead to an information loss, as Solomon has demonstrated. Similarly, when we have a plurality of epistemic systems, we refrain from the pull of dogmatism. However, we can then fall prey to nihilistic relativism if we accept second-order diversity uncritically.
At the very least, arguments favoring the benefit of diversity must specify the kinds of epistemic scenarios to which they are referring. In this manner, cognitive diversity as an epistemic virtue will become ‘sharp’: bound to targeted epistemic practices and deeply context-dependent.18 In other words, there cannot be an a priori benefit of diversity – normative epistemic arguments in favor of cognitive diversity gain their justification entirely from context. To develop a fully operationalizable account of diversity’s benefit, the epistemic arguments explored here must confront epistemic practice itself. For my purposes, this means assessing the history of science from the perspective of first-order and second-order belief-forming. Herein lies the next step of my journey towards sharpening the tool of cognitive diversity.
About the Author
Charlotte Zemmel is an MSc Natural Sciences student at Newnham College, Cambridge, studying History and Philosophy of Science. Her research focuses on social epistemology, philosophy of science policy, and feminist philosophy of science. Her current research is investigating the structure and validity of arguments that encourage the increase of cognitive diversity in epistemic communities. Charlotte Zemmel.
- Stegenga, Jacob. “Robustness, discordance, and relevance.” Philosophy of Science 76.5 (2009): 650-661. p.650
- Rolin, Kristina. “The Epistemic Significance of Diversity.” The Routledge Handbook of Social Epistemology (2019).
- Zollman, Kevin JS. “The epistemic benefit of transient diversity.” Erkenntnis 72.1 (2010): 17.
- Kitcher, Philip. “The division of cognitive labor.” The journal of philosophy 87.1 (1990): 5-22.
- Miller, Boaz. “When is consensus knowledge based? Distinguishing shared knowledge from mere agreement.” Synthese 190.7 (2013): 1293-1316.
- Ibid. p.1311-1314
- Jacobs, Struan. “John Stuart Mill on the tyranny of the majority.” Australian Journal of Political Science 28.2 (1993): 306-321.
- Mill, John Stuart, and Jean Bethke Elshtain. On liberty. Yale University Press, 2003. P.17: “However unwillingly a person who has a strong opinion may admit the possibility that his opinion may be false, he ought to be moved by the consideration that however true it may be, if it is not fully, frequently, and fearlessly discussed, it will be held as dead dogma, not a living truth.”
- Ibid. p.9: “If the opinion is right, they are deprived of the opportunity of exchanging error for truth: if wrong, they lose, what is almost as great a benefit, the clearer perception and livelier impression of truth, produced by its collision with error.”
- Mitchell, Sandra D. “Integrative pluralism.” Biology and Philosophy 17.1 (2002): 55-70.
- Relativism, What Is Epistemic. “Michael Patrick Lynch.” The Routledge Handbook of Social Epistemology (2019): 167.
- Koppl, Roger. “Epistemic systems.” Episteme 2.2 (2006): 91-106.
- Carter, J. Adam. “Disagreement, relativism and doxastic revision.” Erkenntnis 79.1 (2014): 155-172.
- Solomon, Miriam. “Groupthink versus the wisdom of crowds: The social epistemology of deliberation and dissent.” The Southern Journal of Philosophy 44.S1 (2006): 28-42.
- Regina (Daly) v Secretary of State for the Home Department  UKHL 26 para 28, p. 548
- Steinle, Friedrich. Exploratory experiments: Ampère, Faraday, and the origins of electrodynamics. University of Pittsburgh Press, 2016.; Franklin, Laura R. “Exploratory experiments.” Philosophy of Science 72.5 (2005): 888-899.
- Carter, J. Adam. “Epistemic pluralism, epistemic relativism and ‘hinge’ epistemology.” Epistemic pluralism. Palgrave Macmillan, Cham, 2017. 229-249.
- Solomon, Miriam. “Norms of epistemic diversity.” Episteme: A Journal of Social Epistemology 3.1 (2006): 23-36.