owned this note
owned this note
Published
Linked with GitHub
# LSE MSc Dissertation
# On the Ontology of Multidisciplinary Epistemic Groups
###### tags: `LSE`
:::info
* This document charts my plan towards chipping away at the problem I wish to work on in my dissertation for my MSc Philosophy of the Social Sciences.
* To the best of my knowledge, making my thinking publicly available (albeit not findable) does not violate any of the LSE's dissertation regulations.
* If you would like to point out any flaws in the below or would like to simply have a chat about collaboration in academia, please feel free to say hi! You can find me on Twitter ([@hermeneuticist](https://twitter.com/hermeneuticist)) and on [LinkedIn](https://www.linkedin.com/in/ismaelkherroubi/).
:::
[TOC]
# Meetings
## 12 October 2020
Have a gander:
* Fleisher ([2018](https://doi-org/10.1007/s11098-017-0976-4)) on endorsement as our defence of our own theories and conforming to an inclusive notion of epistemic rationality, using models to "not to come down in favor of one solution in that domain, but to show how we can use such solutions in a framework for rational endorsement that smooths the tension between individual and collective epistemic rationality" (p. 2668)
* Heesen ([2018](https://doi.org/10.5840/JPHIL20181151239)) claims "scientists are incentivized to produce more results at the expense of spending more time on the reproducibility of any given result" and uses a model to defend this, arguing "that three basic ingredients are sufficient: first, the fact that speed and reproducibility trade off against each other; second, the fact that scientists get rewarded for publications; and third, the fact that publications depend on peer review, which has to assess the medium-to-long term impact of papers in the short term, and necessarily does so imperfectly" (p. 664)
* Rubin & O'Connor ([2018](https://doi-org.gate3.library.lse.ac.uk/10.1086/697744)) has too much game theory for me...
* Zollman ([2018](https://doi.org/10.5840/jphil201811511)) using basic maths to see how several agents choose between seeking truth, credit or a mix of the two -- I don't think anyone thinks it is either all-credit or all-truth
Concluding: I don't do game theory, but it is interesting to build assumptions into models and see what happens. However, that is where it ends for me: it is interesting, not persuasive.
But could I argue for collaboration along the lines of Fleisher's "endorsement" which is sensitive to both intrinsic (e.g.: truth, consistency) and extrinsic epistemic values?
1. HPS: get a study with failure of collaboration: does Fleisher's approach solve the problem? Either see how it is implemented or how it needs fixing
2. Critique Fleisher if there is a problem
3. Combine the papers: truth-vs-credit seeking in Zollman vs Fleisher's "I want everyone to achieve something collectively"
4. Transitioning into Fleisher's argument, but how do we get there?
Similar to Knorr-Cetina: "Laboratory Life"
See http://www.haixindang.com/research.html and download 2019 "Do collaborators and science..." and email for "Epistemic responsibility in science" mentioning I know Liam. She works on collaboration and thinks a lot about how
* As an HPS-approach, could we look at participatory science and the need for different types of collaboration involved? Not only must scientists from different domains collaborate, but so too must both scientists and participants collaborate to some degree (at the very least, scientists must adjust how they communicate with non-expert participants)
## 19 October 2020
### Dang ([2019](https://doi.org/10.1086/705444)) Do Collaborators in Science need to Agree?
#### Intro (1029-30)
Do we need every coauthor of a scientific article to agree on its conclusion P *as well as* its premises R and Q? No: "collaborators do not, in fact, need to reach broad agreement over the justification of a consensus claim.""
#### Motivating the problem: balancing consensus and epistemic diversity (1030-32)
* *The value of consensus thesis* vs *the value of diversity thesis*
#### Consensus-promoting views of collective justification (1032-34)
The consensus thesis (all must defend P *and* R and Q) is defended by many and of either form:
1. Joint commitment-based collective justification
* A group G is justified in believing that P iff G has good reasons to believe that P; and G has a reason R to believe that P just in case all members of G would properly express openly a willingness to accept R jointly as the group's reason to believe that P. [name=Ismael: this sounds a lot like Gilbert's *joint intentionality*.] There is a clear promotion of consensus over diversity.
2. Aggregation-based collective justification
* Through an aggregation procedure, G's belief in P is justified iff all members' beliefs in P are themselves justified. "In order for a collaboration to be justified in P, the individual reasons why each member supports P must [...] be able to be consistently aggregated." [name=Ismael: Where is this aggregation-consistency requirement from?.] There is a clear promotion of consensus over diversity.
Consensus-promoting views of collective justification don't work because:
* Disagreement is assumed to be irrational
* They do not capture real epistemic processes in successful scientific collaborations
Two justifications for diversity in addition to consensus:
1. "Diversity can be of epistemic value when disagreement and heterogeneity among collaborators is justified"
2. "Diversity of epistemic viewpoints is [...] a fact of scientific practice that should not be taken as a negative feature of social groups"
#### Justified disagreement (1034-36)
Two epistemic "mechanisms" that are sources of diversity of justifiers in scientific collaborations:
1. Multiple sources of evidence [name=Ismael: see Jordan-Young on different theories "fitting"]
2. Different background theories
Consensus wtill matters, but only when designing the research question (at the start) and when agreeing on the conclusion (at the end)
#### Application: Multimethod research in the social sciences (1036-)
* Qualitative methods involving the *causes-of-effects* approach, vs. quantitative methods involving the *effects-of-causes* approach (Mahoney & Goertz, [2006](https://doi.org/10.1093/pan/mpj017))
* Poteete et al. ([2010](https://doi.org/10.1515/9781400835157)) advocate for collaborations that combine both qualitative and quantitative methods
* Case study where computer-based models are confirmed in their findings by empirical data
#### Commentary
Outline:
1. Consensus-promotion: (i) joint-commitment-based and (ii) aggregation-based collective justification
2. Diversity is justifiable too: (multiple sources of evidence and (ii) different background theories)
3. "Collaborators do not, in fact, need to reach broad agreement over the justification of a consensus claim" (2)
My views:
* I feel like this echoes Solomon's understanding of dissent in scientific practice
* I think the Gilbertian notion of joint-commitment-based collective justification also falls on the basis of it being too demanding, as common knowledge implies that I know that you know that I know, etc.
* Is there a moral justification for needing to endorse both R and Q, and P? Would a scientist be happy to have their name on a paper arguing for some agreed-upon P that relies on some R and Q that is not justified in their view?
### Latour ([1979](https://press.princeton.edu/books/paperback/9780691028323/laboratory-life))
#### Chapter 6
Concepts for the argument:
* *Construction*: "The démystification of the difference between facts and artefacts was necessary for our discussion (at the end of Chapter 4) of the way in which the term fact can simultaneously mean what is fabricated and what is not fabricated. By observing artefact construction, we showed that reality was the consequence of the settlement of a dispute rather that its cause" (336)
* *Agnostic*: "If facts are constructed through operations designed to effect the dropping of modalities which qualify a given statement, and, more importantly, if reality is the consequence rather than the cause of this construction, this means that a scientist's activity is directed, not toward "reality," but toward these operations on statements. The sum total of these operations is the agonistic field" (337)
* *Materialisation*: "the same set of intellectual components can be shown to become incorporated as a piece of furniture a few years later" (238)
* *Credibility*: "the synthesis of economic notions (such as money, budget, and payoff) and epistemological notions (such as certitude, doubt, and proof) [and] permits the linking of a string of concepts, such as accreditation, credentials and credit to beliefs ("credo," "credible") and to accounts ("being accountable," "counts," and "credit accounts")" (238-9)
* *Circumstances*: "Our claim is not just that TRF is surrounded, influenced by, in part depends on, or is also caused by circumstances; rather, we argue that science is entirely fabricated out of circumstance; moreover, it is precisely through specific localised practices that science appears to escape all circumstances" (239)
* *Noise* (or the ratio of signal to noise): "information is measured against a background of equally probable events" (240)
Argument: scientists produce order from disorder
Four approaches that converge once we take disorder as the norm and order as the exception:
* "The history of science can be characterised as demonstrating the chain of circumstances and unexpected events leading to this or that discovery. However, this mass of events is not easily reconciled with the solidity of the final achievements" (251). No longer problematic!
* "Sociologists have demonstrated the importance of informal communication in scientific activity. This well-documented phenomenon takes on a new meaning against the newly modified assumption: the production of new information is necessarily obtained by way of unexpected meetings, through old boy networks and by social proximity" (251-2)
* "citation analysts have demonstrated the extensive waste of energy in scientific activity [which] no longer appears paradoxical if we accept the hypothesis that order is an exception and disorder the rule" (252)
* "the diversity of accounts and inconsistency of scientific arguments should therefore come as no surprise: on the contrary, the emergency of an accepted fact is the rare event which should surprise us" (252)
### Knorr Cetina ([1999](https://philpapers.org/rec/KNOECH))
#### 7.2 The Erasure of the Individual as an Epistemic Subjectt
The case of high energy physics (167)
## 26 October 2020
### Huebner & Bright ([2020](https://philpapers.org/rec/HUECRA))
Structure:
1. Scientific fraud is likely to emerge as an effect of partcipating in the scientific credit economy under stress
2. Kinds of collective responsibility at play in scientific communities
3. Problems emerge in large-scale collaborative research and chost-managed research
4. So, who should be held accountable?
4.1. "When scientists act together as members of the scientific community, they can begin to shift the salience of expectations to act in particular ways. And over time this can lead to a shift in the way that the scientific community as a whole operates"
4.2. "When we claim that the community as a whole is accountable foor fraud, what we mean to suggest is that the proper target of praise and blame in the case of fraud is the interconnected network of scientists who act in accordance with their roles as community members, and who shape the salience and credit-seeking norms"
### Huebner et al. ([2017](https://doi.org/10.1093/oso/9780190680534.003.0005))
Research architectures:
1. Catch-and-toss authorship
2. Centralised control authorship
3. Radically collaborative research (which unfortunately cannot sit under either of the other two (104-108))
### Patternote ([2020](https://https://doi.org/10.1007/978-3-030-29783-1_3))
Social ontology literature tends to focus on cases of *minimal* joint action, but there are important cases of *maximal* joint action that are worth studying:
1. Demonstrations: due to *epistemic obstacles*, "it must be reached by means that are not at work in other joint actions"
2. Deliberation: *strategic obstacles*
2.1. Goals are not clearly defined in terms of component actions
2.2. Although extended in time, the usual dynamical conditions for joint action do not apply
2.3. Deliberation clashes even with the bottom-up, mechanistic approach to joint action because we have good reason to believe it amplifies the negative effects of individual biases
3. Collective free improvisation: "a musical performance that ia accomplished without any underlying background musical work by performers who may have never played together before and favour different styles. [...] It does not just make coordination difficult by multiplying the coordination options, but also because *standards for a successful coordination may not be shared by the performers*."
There are four ways minimalism manifests in social ontology:
1. Scale
2. Conceptual
3. Ontological
4. Cognitive
### Notes: 26 October 2020
Email LKB by next Monday:
1. Write a proposed essay structure: what's the thesis (see maximal instances of joint action) and supporting reasons
2. Choose "a case" or "collaborative level"
3. Or problems
## 02 November 2020
* No need to argue for the importance of the social ontology of these groups
* ASAP action: write down what the problems of joint actions are, and the kind of questions in the
* 500-1000 words about the problem of
* Next up: 16 November
## 16 November 2020
* Keep going!
## 21 November 2020 (with the seminar group)
*Questions:*
* Are you interested in how groups know or how we create knowledge as groups?
* The latter
* How does intersubjectivity fit in your account?
* Not enough, I want more intersubjectivity now!
* Are you interested in how individuals' knowledge is shared with *other individuals* or *amongst groups*?
* The former, although as a barrier to the epistemic (scientific) success of MEGs.
*Recommendation:* JAM suggested looking into Birch's "Joint Knowledge" and Zollman's "Zollman Effect".
## 30 November 2020
Wilholt (2016):
1. Value judgements shape one's preference for false negatives over false positives or vice versa.
2. Therefore, value judgements affect the methodology chosen for scientific inquiry.
3. Scientific communities have shared methodological standards which constrain the choice of method.
4. Researchers needn't guess at peers' value judgements when constraining methodological standards are used.
5. A harmonisation of of methods is necessary for epistemic trust within scientific collaborations.
6. Therefore, value judgements and decisions about methods are made at the group level, rather than by each individual (231).
Questions for Liam:
* In arguing that MEGs are metaphysically distinct from their constitutive members, can I use "irreducible" as shorthand?
Liam suggested Poteete et al.'s' (2009) *Working Together*. Upon inspection, the refined IAD framework (58-59) might help elucidate the conceptual maximalism of MEGs. In terms of ontological maximalism, it is yet to be seen how MEGs rely on irreducibly social conventions.
## 07 December 2020
The correct reading of ontological maximalism in multidisciplinary epistemic groups (MEGs) is that there are judgements made by MEGs that are not reducible to individuals' judgements.
* But couldn't it be that all the individuals agree with the MEG's judgement *P*?
* The claim is not that the MEG judges *P*, but makes some value judgements and sustains some premises *R* that are not reducible to its members' beliefs. Given this caveat, the response is in the negative on two bases:
1. The value judgements are only *implied* by methodological standards. The question above requires that each individual both (i) agree with the standard's *implied* value judgements, and (ii) be *aware* of such judgements. The combination of these two factors is unlikely. Although I grant that the case of researchers who reflect on the ethics underlying the methodological standards they employ is possible, the above question requires that *all the individuals agree* and meet both conditions.
2. In MEGs, there are no shared methodological standards, so (1) -- if it holds -- will only be true amongst those of a certain field within the MEG. Each premise *R* that upholds the conclusion *P* requires a certain training and expertise. The MEG's premises *R* cannot constitute a justified true belief of each researcher in the team.
## 11 January 2021
Conceptually maximal MEGs; thesis: academic incentives heavily influence individual researchers' plans, which, in turn, are constitutive of the nature of MEGs. Any account of the ontology of MEGs, therefore, requires an account of various academic incentives. This is in line with Bratman's account of "shared intentionality", which relies precisely on individuals' intentions as defined by his planning theory.
I argue that these intentions ought to be understood as plan states following Bratman's conceptually maximal planning theory. I rely on Bratman's mention of "research teams" to illustrate this (Bratman, [2014](https://global.oup.com/academic/product/shared-agency-9780199339990?cc=gb&lang=en&): ch. 7).
Not congruent: necessary aspect of group is a shared plan, but this is unlikely,
No shared plans; is Bratman happy with this?
There is a literature on norms as social expectation -- we are not engaging with it and must not speak of "academic norms".
As an example, Bratman does speak of deliberation in the context of "research teams" that show, therefore, shared intention. Although, this is not on the basis of shared value judgements, but a commitment to shared intentions.
---
## Notes: 04 April 2021
Multidisciplinarity accommodates for the reference class problem
Problems highlighted by conceptual maximalism can be counteracted by the sense of unity brought about by ontological maximalism
---
## Notes: 26 January 2022
* Clear this HackMD using [this one](https://hackmd.io/P2y4ZP2fR6CjwJ2TWZqyjA?view) to capture old notes and resources.
## Meeting AL: 03 February 2022
We have an intuition about MEGs whereby they have some plural agency, but can this be vindicated in light of joint action theories that deal with much simpler groups?
1. Start with the difference between a duet and a MEG.
2. Explain the problems theories face, and explain how they need modifying.
3. How do we get around sub plans not meshing?
4. Challenge to the practical contribution: why is joint intentionality helpful?
## Seminar presentation: 04 February 2022
* Number of concepts or *information density*
* Conceptual maximalism or *explanatory specification*
## Notes: 22 February 2022
* Intro to literature starts with Philosophy and sociology of science requiring joint action theory, and then joint action theory being complemented by P/SoS because it is, otherwise, too minimalist.
* Following section is, basically, on ontological maximalism.
## Notes: 24 February 2022
* Wrote §Socialising Science
## Notes: 16-Mar 2022
Bratman speaks of a "continuity constraint" — his theory of modest sociality and shared intention assumes continuity from the planning theory of individuals to the shared intentions of groups. Bratman points out that Gilbert's theory is one of "discontinuity".
"Shared commitments to weights" (135) of some goal over another (there is an example of the agreed weighing that research members might share, for example on the need for their work to result in some publicly beneficial output, versus the design of lucrative patents).
Policies about share intentions about weights are described on page 1410 is 2014 book.
## Meeting AL: 17-Mar 2022
* is the point of Bratman's norms to attribute some sort of rationality to "shared intentions"?
## Meeting AL: 24-Mar 2022
Epistemic humility in regards to the limitations of a discipline, not only myself. I trust my discipline, this colleague says to do it another way. Humility for the discipline is cool.
The problem is we can’t agree on meshing sub plans.
Different disciplines have different values (e.g.: methodological standards)
“Status quo bias” >> “do what you are comfortable with” worries
Epistemic humility can mean taking another’s values seriously (compromise, or adjust your view).
## Notes: 22-Mar 2022
I’ll have better defined questions about the links between (i) the challenges multidisciplinarity raises, (ii) the “norms for modest sociality” Bratman defines, (iii) and the policies/values such teams can draw on to overcome the identified challenges.
| Challenge | Bratmanian norm | Values-based policy | Order |
| --- | --- | --- | --- |
| --- | Agglomeration | Epistemic transparency to acknowledge the role of academic norms in shaping peers' broader goals beyond a project | 2 |
| Training-expertise trade-off (you are incentivised to specialise and work in narrowly defined fields) | Consistency & coherence | Social epistemic humility to encourage deliberation about alternative courses of action | 3 |
| --- | Stability | Medina's open-mindedness | |
| Reference class problem | --- | Outsource ultimate decision about publication venue to institution's governance frameworks | 1 |
| --- | --- | | |
**Agglomeration** is a solution insofar that it helps acknowledge different non-intentional goals that are practically irrelevant to shared intentions. **We need a policy that captures this insight.** Awareness of the endogenous influences that are academic norms can at least help this agglomeration process be more transparent. [*Epistemic transparency*?]
**Coherence and consistency** inspire Bratman to introduce – albeit superficially – *deliberation*.
>The values of the individual make it possible to consider alternative courses of action and arbitrate between them. Likewise, a team’s shared values will provide the backdrop for productive group deliberation. Disagreement will provide a way of assessing whether sub-plans mesh, rethinking shared values, and fine-tuning the team’s plans (Tollefsen, [2006](https://doi.org/10.1353/epi.0.0008): 44).
Tollefsen makes me think that (i) deliberation respect for a set of "core values" against which dissent would be unacceptable (44-45), and (ii) people should enter MEGs with a mind open to hearing of other avenues of action, which sounds like Medina's ([2013](https://doi.org/10.1093/acprof:oso/9780199929023.003.0001)) open-mindedness (44).
**Stability** is about seeking to successfully mesh subplans throughout time. If my following claim is faithful to Bratman:
>The snowball effect becomes social when we acknowledge this need for meshing sub-plans with others.
Then we have Medina's open-mindedness again. Social epistemic humility
## Notes: 23-Mar 2022
~~**Social agglomeration** inspires needing to be aware that peers have lives outside of the lab – broader goals. For example, one might prefer to publish in one venue over another. I suggest **academic institutions "absorb" the responsibility of identifying adequate venues for publishing** the work they are funded to do.~~
> [name=Ismael-KG]MEGs might often be pan-institutional, maybe the funder decides (e.g.: UKRI strongly recommends ). Forget it.
~~Such a preference can influence one's "credit" in different ways according to one's career stage. I **suggest transparency about academic stages** and "otherly goals" inform decisions that institutions make. With this, we have a policy in line with Bratman's **social agglomeration**.~~
Making such broader goals clearer can also help see that one's peers partake in the same academic norms (an important assumption discussed when overcoming methodological differences above). The courses of action (methodologies) proposed from other epistemic backgrounds, then, can be granted credibility by virtue of their being rooted in said shared norms. Bratman's **consistency and coherence** can be attained by embracing a **social epistemic humility**, or the idea that one's specialised training does not detract from another's know-how.
> [name=Ismael-KG]Adam likes the last bit. Focus on the norm of epistemic humility anddescribe how it works in ordinary scenarios before building to what it means in MEGs. Shared academic norms is insufficient, so spell out epistemic humiity very clearly.
This humility can be complemented by what Medina calls **epistemic open-mindedness**, whereby one is open to perspectives that might otherwise "destabilise (or create trouble for) one's own perspective" ([2013](https://doi.org/10.1093/acprof:oso/9780199929023.003.0001): 35). This goes further than acknowledging one's field's limitations by encouraging proactive engagement with others' perspectives and expertise.
> [name=Ismael-KG]How often do I look for disagreements? Do I ask for your evidence? Humility is about reconciling new evidence from others, but we can go further by looking for disagreements. Finding disagreements
## Notes: 29-Mar 2022
**Abstract submitted** (via PH499 on Moodle; word-limit: 100)
>Research teams can be conceived of as plural agents. In this essay, I apply joint action theory to the case of multidisciplinary epistemic groups (MEGs); research teams whose members come from different disciplines. Specifically, I study MEGs that operate in academic institutions. I argue that how trust is attributed to MEGs, as well as the way they make certain judgements, renders them as ontologically distinct from their individual members. I engage with Bratman’s planning theory of action to test this view. Bratman's increased explanatory specificity points to challenges for MEGs. I briefly suggest measures for MEGs to overcome these challenges.
## Notes: 01-Apr
The norms are for individuals to be rational in constituting cases of modest sociality.
Agglomeration is ill-defined by Bratman and refined by Zhu ([2010](https://www.jstor.org/stable/40664939)) who explains agglomeration as the ability to form compound intentions that are co consistent with previous intentions and allow for "efficiency" insofar that we can focus on "menial" tasks without having to recollect broader and more complex compound intentions at each stage.
### Meeting with AL
On Wilholt, maybe we don't need to attribute trust to groups, but it helps us get to the truth more easily – the idea that attributing trust to groups is helpful, more efficient.
For the final section, do this:
1. Define epistemic humility (800 words)
2. Give an ordinary example (we disagree about who owes what in a bill, and my confidence drops because I am humble; 300 words) – this is it, but we can go a step further by suggesting a methodology we agree on to establish the truth)
3. Build up to a MEGs example: keep details vague but explain what the differences are with the ordinary example (my confidence drops because "my field makes mistakes", **NOT ME**)
4. We can go a step further by suggesting a methodology we agree on to establish the truth. The method to establish this methodology is beyond the scope of this paper.
## Notes: 07-May
:::info
wordcount: 1,076
:::
---
# Dissertation (Feb 2022)
Research teams can be conceived of as plural agents. Social ontologists such as Gilbert have argued that some social groups can exist as entities that are ontologically distinct from their individual members. She speaks of "plural subjects" (Gilbert, [1989](https://doi.org/10.2307/j.ctv10vm20z.7)) when describing groups of people who share certain commitments and see themselves as part of a group. However, an equally influential scholar in the field has advocated for the opposite case. Bratman ([2007](https://doi.org/10.1093/acprof:oso/9780195187717.001.0001)) has argued that irreducible group intentions don't exist. Rather, a social group can be sufficiently explained by a theory of action that describes its individual members; joint intentions are reducible to patterns of individual intentions.
Bratman's conclusions contrast with some philosophy of science literature. At least two accounts of scientific practices better align with Gilbert's "plural subjects". On the one hand, Wilholt has argued that researchers attribute trust to research groups rather than to their individual members (Wilholt, [2016](https://doi.org/10.1093/acprof:oso/9780198759645.003.0012)). By his account, peers working in the same field can trust the quality of work of research groups by virtue of their adhering to certain methodological standards. On the other hand, research groups have been found to have the capacity to make judgements. Dang ([2019](https://doi.org/10.1086/705444)) proposes a framework to see how researchers might come together in developing scientific conclusions as parts of "plural subjects" *without* needing to agree with one another's methods or premises.
In this essay, I scrutinise these conflicting views in the case of multidisciplinary epistemic groups (MEGs); research teams whose members come from different disciplines. Specifically, I am interested in MEGs that operate in academic institutions. I argue that Wilholt's trust-attribution account and Dang's argument for group-level judgements render Gilbertian "plural subjects" a plausible description of MEGs. I engage with Bratman's planning theory of action to test this view and suggest its utility for making sense of MEGs. However, the increased explanatory specification Bratman points to concerns for the feasibility of multidisciplinarity in academia. By drawing on literature on epistemic virtues, I briefly suggest measures for MEGs to overcome these challenges.
## Navigating the Literature
In this section, I introduce the two philosophical sub-fields that motivate this essay: social epistemology of science, and joint action theory.
### Socialising Science
By social epistemology of science I mean the study of knowledge-production as a social endeavour. The field can be traced back to at least as far back as John Stuart Mill's work in the nineteenth century (Longino, [2019](https://plato.stanford.edu/entries/scientific-knowledge-social/)). Mill ([1859](https://socialsciences.mcmaster.ca/econ/ugcm/3ll3/mill/liberty.pdf)) identifies the need for dialogue in the pursuit of justifying beliefs and reaching consensus as to what is "true". Agreeing upon what is true as a collective process mitigates individuals' fallibility and cognitive limitations. The framing of individuals' fallibility would inspire Popper's theory of falsificationism, whereby science proceeds by identifying the limitations and inaccuracies of contemporary scientific theories (Popper, [CITE]()).
Even greater emphasis on the social nature of knowledge production was given by Kuhn, whose historical analysis of scienttific developments garnered insights about how "normal science" precedes "revolutions" that lead to "paradigm shifts" (Kuhn, [CITE YOU-KNOW-WHAT]()). To take one reading of his analysis, paradigms are the sets of beliefs, tools and methods that shape scientific inquiry during periods of "normal science". Whilst Mill and Popper had already pointed to the dialogical process of science as a collective endeavour, Kuhn identified the importance of analysing the history of science and how social contexts changed and, therefore, changed scientific methodologies.
The social epistemology of science has since gained from methodologies employed by scholars such as Merton, Latour and Knorr-Cetina. These scholars, amongst others, deployed anthropological methodologies to the study of how scientific projects are practically conducted. Merton ([CITE]()) is often cited as the founder of the sociology of science. Merton applied sociology to the study of scientific practices. One key insight was the use of studying incentive structures and norms within academic institutions to better gauge the processes of knowledge-production. This strand of literature will be crucial in my analysis of the practical implications of joint action theory as applied to the study of MEGs. Before introducing that second area of literature, I want to clarify the term "multidisciplinarity."
Mäki ([2016](https://doi.org/10.1007/s13194-016-0162-0)) provides calls for consensus-building in what he calls philosophy of interdisciplinarity. One reason for this is the growth of interdisciplinary approaches to research question in recent decades (see Van Noorden, [2015](https://doi.org/10.1038/525306a); Earnshaw, [2020](https://doi.org/10.1007/978-3-030-42097-0_20)). Crucially, Mäki identifies heterogeneity in forms of interdisciplinarity, which might respond to the nature of the different disciplines being brought together, or the aims they strive for (Mäki, [2016](https://doi.org/10.1007/s13194-016-0162-0): §5). To avoid making claims about interdisciplinarity that are only narrowly applicable, he conceives of it "in terms of whatever relevant relationship between two or more scientific disciplines or their parts" ([*ibid.*](https://doi.org/10.1007/s13194-016-0162-0): 331). I wish to do the same but use the term following Jantsch and Piaget's "three levels" of interdisciplinarity, whereby "multidisciplinarity" is less demanding on the relationships forged between the disciplines than "interdisciplinarity"; and "interdisciplinarity," in turn, is less demanding than "transdisciplinarity" (OECD, [1972](https://eric.ed.gov/?id=ED061895): 133-136)[1: Although it must be noted that there is little consensus about these terms (see Lawrence, [2010](https://www.ed.ac.uk/files/imports/fileManager/RJL-2010Inter-Trans.pdf))]. Jantsch and Piaget suggest multidisciplinarity to emerge where "persistent heterogeneity of information used" is present in research teams with "interdisciplinary objectives" ([*ibid.*](https://eric.ed.gov/?id=ED061895): 133). Thus, I propose the following working definition of multidisciplinarity for the present study of MEGs:
> *Multidisciplinarity* is present in research teams where there is heterogeneity of methodological standards.
### Joint Action Theory
The study of social groups has long and complex history in philosophy. For example, we commonly attribute the conception of humankind as a social animal to Aristotle. However, it is only through chance that this is the case. As Arendt ([1998](https://www.goodreads.com/book/show/127227.The_Human_Condition): ch. 4) explains, it is actually down to a series of translations that have left us with Aquinas' interpretation that "humankind is by nature political, that is, social" rather than Aristotle's focus on humankind as *political*. We can also see in Hobbes' ([1651]()) work the concept of *covenants* as playing a role in the construction of the social world. Puffendorf ([1673]()) later speaks of *conventions*, and Hume ([1740]) expands on these with social concepts like *money*, *government*, and *promises*. But at no moment do we find a framework that describes what people actually *do* to constitute the social world. This changes with Émile Durkheim and Max Weber entering the scene in the late 19th century. Durkheim ([1994](http://www.worldcat.org/oclc/29386457)), on the one hand, sought to understand how individuals stand in relation to customs and conventions, or "*faits sociaux*", which he found to have "coercive power" on people, who inevitably participate in society. Weber ([1978](https://doi.org/10.1017/CBO9780511810831.005)), on the other hand, sought a framework to understand what made human actions social, proposing that directing the intention of one's action towards another rendered it a social one. Thus, social ontology presents us with a myriad of challenges: on individuals' intentions, their relations with conventions, and the forming of *social groups*.
Having defined multidisciplinarity in the previous section, I am interested in how theories on social groups might be deployed to explain MEGs. I suggest seeking a relevant account in the philosophical tradition of joint action theory. Very roughly, we can see joint action as those actions that involve various agents. An example of joint action can help discern what is being discussed here.
Consider Bratman's ([1993](https://www.jstor.org/stable/2381695)) study of a musical duet, two people who succesfully sing together. Joint action theory is, then, about identifying the conditions that allow for such a duet to perform together. Questions in joint action theory ask whether the two singers share rules for action (perhaps they sing in a certain key), share intentions (maybe to stun an audience), must share understandings of one another's aims (that is, make their aims "common knowledge"), and so on.
There are two prominent and opposing views in joint action theory. Michael Bratman is on one side of the intellectual battle. On the other side is Margaret Gilbert. Let us briefly see what the two argue for and where some of their differences lie.
To start with, Gilbert ([1989](https://doi.org/10.2307/j.ctv10vm20z.7)) provides an account of social groups as *plural subjects*. Without providing an exegesis of her work, two key factors grant a social group the status of plural subject: how one uses the word "we", and *joint commitment*. Firstly, for there to be plural subjects, the term "we" has a very specific meaning, as it refers to "a set of people, each of whom shares with oneself in some action, belief, attitude, or other such attribute" (201). Gilbert argues that "those who constitute a plural subject know that they do, and will thus think of themselves as *us*" (205). Secondly, the individuals of a group must be "jointly committed" to act and believe *qua* social group. This joint commitment must be expressed in some way by each individual; although it can be present implicitly, such as when walking together with a friend (Gilbert, [1990](https://doi.org/10.1111/j.1475-4975.1990.tb00202.x)). Let us consider this very case. In two people walking together, the very act of walking side by side and at the same pace shows that each individual is committed to the goal of walking in one another's company. In sharing this commitment, what results is a *plural subject*, as they can use "we" in the way mentioned earlier; they can say “Shall we stop here?” or “Shall we go through the woods?" (8). Whichever way the joint commitment is expressed, it must be known to fellow group members (i) that it is the case that all other group members express such commitment, (ii) that each group member is aware of fellow group members' commitment, and (iii) that this form of *common knowledge* is shared amongst the group in what regards their joint commitment (Gilbert, [1989](https://doi.org/10.2307/j.ctv10vm20z.7)).
Summarising the above, a plural subject exists when the members of a group have common knowledge of each others' joint commitment and, therefore, can speak of "us" in Gilbert's sense. One example that Gilbert provides is that of a poetry discussion group ([1987](https://doi.org/10.1007/BF00485446)), weherein a plural subject is formed on the basis of the group being able to decide on an interpretation of a poem without the need that any group member believe in this interpretation. What Gilbert ultimately provides is a framework to understand individual persons as distinct from the social groups they form a part of, as these groups have beliefs and intentions of their own. I will later refer to this as *ontological maximalism*.
Turning to Bratman ([1993](https://doi.org/10.1086/293577)), we find a more individualistic account of social groups, whereby each individual of a group has an intention of the form "I intend that we *J*". Each fellow group member is aware of everybody else having this intention. Much like in Gilbert's account, there is a role to be played by common knowledge. However, the "collectivist" aspect of Gilbert's account is lost. For Bratman ([2007](https://doi.org/10.1093/acprof:oso/9780195187717.001.0001)), irreducibly social intentions do not exist. Indeed, Bratman's account takes individuals' mental states to provide sufficient conditions for "shared intentionality". Thus, he develops a "planning theory" wherein the focus remains on individual intentions (Bratman, [2014](https://doi.org/10.1093/acprof:oso/9780199897933.003.0001)).
By Bratman's "planning theory", intentions go beyond immediate actions. Indeed, intentions play an organising-coordinating role, whereby one conceives complex goals and the cross-temporal preliminary steps that are embedded within. Bratman calls these "plans" and "subplans", respectively. Bratman associates this planning role with norms of intention rationality. For intentions to be rational, Bratman argues, they must meet the following criteria: (i) be internally consistent with one’s beliefs, (ii) be consistent with agglomerated versions of these intentions, (iii) be coherent with the plan’s stages and goals, and (iv) tend towards stability throughout time.
Whilst Bratman and Gilbert are in opposition, both their theories of joint action are limited in their description of "minimalist" characterisations of "simple, intuitive cases" (Paternotte, [20XX](): 42). In the meantime, MEGs are intuitively quite complex. As described above, various disciplines are coming together in research groups that contain more than two researchers, as well as links to governance and administrative support provided, sometimes, by various reseach institutions [**is this already mentioned above???**]. It is for this reason that this essay complements joint action theory with the philosophy and sociology of science literature outlines above.
In what follows, I narrow the scope in regards to philosophy of science literature by drawing on Dang's ([]()) theory of group-level judgements within research groups, and Wilholt's ([]()) theory of trust-attribution to research groups as a whole.
## Gilbertian MEGs
In this section, I introduce two arguments in the philosophy of science literature that support a Gilbertian notion of MEGs, whereby MEGs exist *qua* groups; as something distinct from their individual members. I then conclude that both these arguments align with an example of ontologically distinct groups Gilbert provides when discussing a poetry reading group.
The two arguments are substantively different. The first speaks of ascribing trust to groups *qua* groups.
Consider how we attribute agency to businesses in conversation. I might say "the organisation X's actions are detrimental to the environment" and conclude "X doesn't deserve my money" (see List & Pettit, [2011](): §). Here, I would be *ascribing* agency to a group. I am assuming *X* makes decisions *qua* *X* and that my decisions (to no longer buy from them) will impact them. I call this the *attributional argument for ontological maximalism* (AFOM), and it will be made in the case of MEGs by Wilholt ([2016](https://doi.org/10.1093/acprof:oso/9780198759645.003.0012)).
> [name=Ismael-KG]Awaiting [Marabel's response](https://twitter.com/marabelceline/status/1497561496229871620?s=21). This will determine whether the distinction is at all useful (i.e.: whether this and the following paragraph are helpful or not).
The second argument is for the "intentional states" (List & Pettit, [2011](https://doi.org/10.1093/acprof:oso/9780199591565.003.0002): 21) of agents. Rather than *ascribing* properties to groups, this argument proposes that groups that meet certain conditions [**WHAT CONDITIONS? WILL THIS BE EXPLAINED?**] *have* "intentional states." I speak of *intentional states* in the sense of proposing conclusions and making value judgements *qua* groups. By building on Dang ([2019](https://doi.org/10.1086/705444)), I suggest MEGs have intentional states and, therefore, exist in the ontologically distinct sense Gilbert proposes.
### MEGs' decisions: from value judgements to methodological standards
Beliefs are held by subjects. Trust in those beliefs depends on the trustworthiness of the subjects that hold them. As we will see, trust in MEGs depends on the trustworthiness of the judgements they make *qua* groups. In the case of MEGs, we speak of *epistemic trust*. Consider how researchers build on one anothers' work within their field. When doing so, they must place epistemic trust on others' findings. In MEGs, this epistemic trust is not reducible to their individual members. This can be argued along the lines of traditional social epistemologists, who claim that groups show greater competence than the sum of their parts (see Solomon, [1994](https://doi.org/10.2307/2216062), [2006](https://doi.org/10.1111/j.2041-6962.2006.tb00028.x); Tollefsen, [2007](https://doi.org/10.1080/02691720701674163)). Researchers trust MEGs as groups on the basis that these are better suited to build knowledge than any individual -- or so the argument would go. However, I wish to pursue an argument outlined by Wilholt ([2016](https://doi.org/10.1093/acprof:oso/9780198759645.003.0012)), whereby trust in scientific collaborations relies on their own social conventions. These social conventions, in turn, rely on "how trust *within* the group is enabled and maintained" (Wilholt, 2016:220). To understand these conventions, I first describe what is implied by them: value judgements.
Setting aside the epistemological discussion Wilholt provides in what relates to the reliability of one approach to a scientific question over another, I will speak of "methodologies" *simpliciter*.
The choice of one methodology in a scientific project over another can rely on value judgements. As an example, value judgements manifest in one's readiness to generate a study that is prone to either false positives or false negatives. Consider the ongoing COVID-19 pandemic. For the purpose of tracking the epidemiological evolution of the virus, tests were created to a show whether or not people were carriers of the virus. Unfortunately, such tests cannot be infallible. There is a chance that any one test shows either that the patient has the virus when they do not -- this is a false positive --, or that the patient does not have the virus when they effectively do -- a false negative. In the context of the pandemic, a positive test could mean having to quarantine for a period of time, an action which -- on a large scale of many positive cases -- might result in an economy's contraction. A negative test, in turn, allows the patient to continue going out but risks spreading the virus further, as well as more deaths, in the case of false negatives. Given that the test cannot be accurate 100% of the time, epidemiologists must decide in their studies how to treat data resulting from these tests. Should they allow a wider or smaller margin of error that allows for more false negatives or more false positives? The trade-off is, by this brief outline, between the economy or livelihoods on the one hand, and public health or deaths on the other. This is by no means an easy decision, and science does not provide an answer here. Thus, we turn to value judgements. These, as we see below, are shaped by *methodological standards*.
In practice, scientific communities share "methodological standards". Following these standards allows for scientists to place epistemic trust in scientific findings. An example methodological standard is holding 0.05 as the p-value, the highest acceptable significance level in significance tests. Applying this standard to the above example would limit the list of methodologies to choose from. Epidemiologists would seek methodologies that aim for a p-value below or equal to 0.05. By constraining the choice of methodologies, methodological standards carry implicit value judgements. Setting 0.05 as the p-value shapes our tolerance to false negatives and fakse positives. What's more, methodological standards are social conventions, which means that epistemic trust relies on irreducibly social entities. Epistemic trust is conferred to scientific collaborations that employ methodological standards and *not* their individual researchers. The upshot of this is that value judgements and decisions about methods are made at the group level, rather than by each member.
#### The problem of methodological standards
This interpretation of Wilholt (2016) would suffice to demonstrate ontological maximalism in MEGs. However, *MEGs do not have shared standards*. I will argue that this is not problematic. One strategy to adapt Wilholt's account to the context of MEGs might be to use a weaker notion of "methodological standards". Consider some of the methodological standards he mentions: the use of positive control groups in the case of animal testing, or the registration of of clinical trials with a public repository in biomedical research (222). These are rather field-specific cases of methodological standards. They will not work in multidiciplinary enterprises. Therefore, we need a weaker form of such standards. The case of the p-value discussed earlier on is also provided by Wilholt and can be shared by many more fields. Indeed, significance tests can be employed in any scientific project that involves some statistical analysis. However, there is still a concern in the context of MEGs that this specific standard is not relevant to all the fields that come to form the group. Consider the case of digital humanities, where historians may not employ significance tests at all. What we are after is a standard that is relevant to all scientific fields. This is unlikely to exist, so the strategy to find a weaker notion of methodological standard is not straightforwardly viable.
The strategy I follow is to focus not on methodological standards *per se*, but on the factor that render these useful in describing scientific communities as irreducible to their individuals. Indeed, Wilholt explains that "they are conventional in character and, thus, irreducibly social" (223). So, what we seek is a social convention that is present across research fields. This is a straightforward task thanks to our fixing the institutional context of MEGs: academia. This will be expanded on in the following section on conceptual maximalism within MEGs. For now, what we should note is that the institution of academia carries a series of social conventions. With MEGs enacting these social conventions, we needn't rely on Whilholt's talk of methodological standards, but his argument that social conventions allow scientific communities to make decisions and value judgements *qua* collectives.
#### Embracing epistemic diversity
We have so far seen that, conforming to social conventions, MEGs become ontologically distinct from their members. Having shed the dependence on methodological standards for my account of MEGs, I wish to turn to Dang ([2019](https://doi.org/10.1086/705444)), who does not presume any shared methodological standards. As Dang explains, diversity in scientific practice is a matter of fact. This is even more true of MEGs, where epistemic diversity is a constitutive factor. This diversity results in researchers from different backgrounds having different methodologies when approaching scientific questions. Disagreements about methodology, in this respect, are to be expected. The question then rises: how can MEGs agree on their collaborations' outputs? To respond to this, Dang speaks of a collaboration's conclusion *P* and its premises or reasons *R*. Given the diversity in MEGs and the inevitability of disagreement, Dang argues for a model of collective justification whereby *P* needn't require that all members of the collaboration agree with *P* but that the group *reach consensus on P*. Whilst any disagreement over *P* needs justifying, the reasons *R* needn't be shared by collaborators. This is precisely because of the lack of shared methodological standards.
As an example of a MEG wherein reasons *R* are too diverse to require agreement from every researcher in the group, consider McGillivray et al. ([2019](https://doi.org/10.1093/llc/fqz036)). In their research, they sought to apply computational models to track the evolution of the meaning of words in Ancient Greek texts throughout time. This required rather distinct sets of skills and knowledge We can point to these by defining the project in three stages. Firstly, there was the need to identify a corpus of Ancient Greek texts for their semantic analysis. This was The Diorisis Ancient Greek Corpus (Vatri & McGillivray, [2018](https://www.doi.org/10.6084/m9.figshare.6187256)). Secondly, the team's clacissists annotated a selection of sentences that contained three polysemous lemmas (*harmonia*, *mus* and *kosmos*) used frequently throughout the corpus. The resulting dataset can be found in McGillivray ([2019](http://doi.org.gate3.library.lse.ac.uk/10.6084/m9.figshare.c.4445420)). Finally, the team provide an analysis of the application of two computational models to the annotated texts. These are SCAN (Frerman & Lapata, [2016](https://doi.org/10.1162/tacl_a_00081 )), which applies a Bayesian model to the case of semantic change in English since the eighteenth century; and GASC (Perrone et al., [2019](https://www.aclweb.org/anthology/W19-4700.pdf)), which seeks to help with the scaling up of pattern-detection in the semantic evolution of the Ancient Greek lexicon.
With this project, we find a co-dependence between two groups of field experts -- let us call them "classicits" and "computer scientists". Whilst it is true that both groups would agree with the final outcome of the project -- the foundation for a computational model to track the meaning of words in Ancient Greek --, there is no requirement for both to agree with the other group's methods and premises. Consider the difference between one's methodology and the other's, as well as the expertise required at each stage. It is not feasible to expect that each group -- let alone each researcher -- comprehend the work conducted by the other. With this, it is not required that each researcher agree with each others' premises. To this effect, we have an account of scientific collaboration without methodological standards. What's more, we have some output *P* made by a group versus some premises *R* held by individual researchers. The point here is that there is an irreducibly social group that makes judgements *qua* group.
#### The poetry dicussion group
To relate the above account back with Gilbert's account of group ontology and clarify this section's claim, consider Wilholt's (2016) conclusion that the implicit value judgements embedded in methodological standards are binding on a community's researchers "in the sense that they have to act as if they endorse them when they perform certain research-related activities" (230). There is a supervenience of the MEG's judgements upon its constitutive researchers. The researchers are prepared to *believe X* or *enact X* in the way that the MEG does. Joint commitment, as introduced earlier, requires that each individual do what they can "to emulate, as far as possible, a single body that Xs" (Gilbert, [2013](https://doi.org/10.1093/acprof:oso/9780199970148.001.0001): 176). Gilbert provides an example of a group believing in a way that needn't replicate in its members' beliefs. In "the case of the poetry discussion group" (Gilbert, [1987](https://www.jstor.org/stable/20116447): 190-194), we are asked to consider how such a group holds its meetings. To start, a poem is read out. Afterwards, group members can comment on the reading, offering suggestions and solving any opposing views through discussion. A preferred reading eventually emerges, the discussion comes to a close and the poem is read out once again; this time in line with how the discussion evolved. At this point, it is natural for members of the discussion group to make claims such as "*we* found the final line quite moving". The group, as a whole, is ascribed a belief. This ascription is adequate on the basis that all or most members decided to let the final reading "stand" as that of the group. There is no need, in other words, to accept a certain reading *as an individual*, but to be prepared to participate in a group's acceptance of some proposition. Using Dang's nomenclature, this means that we cannot ask each member of a MEG to believe in every *R* that sustains some scientific output *P*. With all this, I take it that MEGs are not reducible to their constitutive researchers and their philosophical study requires an Gilbertian account of their ontological status.
Whilst the arguments Wilholt and Dang provide sustain Gilbert's view, a great deal of work in the sociology of science might put it against the ropes and require greater explanatory density. I propose, in the following section, that Bratman's individualistic planning theory of action aligns with the complex nature of the academic ecosystem.
## Bratmanian Explanations
In this section, I demonstrate Bratman's theory is relevant in describing the ontology of MEGs. Firstly, I briefly survey the academic structures of universities. In particular, I describe how “academic incentives” feed into a culture of “publish-or-perish”. This culture, it will be shown, emphasises the importance of individualist goals in MEGs. Secondly, I introduce two challenges that MEGs face due their multidisciplinary nature. Finally, I suggest Bratman's framework for "modest sociality" helps explain how divergent plans within MEGs can mesh such as to overcome these challenges.
### Academic Norms
I will speak of "academic norms" to refer to those conventions that proliferate in the context of research institutions -- universities, more specifically. We have already seen how social conventions that shape scientific inquiry render MEGs irreducibly social. This section introduces two interrelated factors that constitute academic norms: the reward structure and the peer review process. Whilst I do not intend to provide an exhaustive account of academic norms, it will be argued that these two norms feed into what is commonly denoted the "publish-or-perish" culture. This is a much wider concept that calls for conceptual maximalism, as we will see in the following section.
The reward system in academia shapes how individual researchers decide on the projects to pursue and the areas to train in. This can be seen by considering what Zollman (2018) calls "the credit economy in science" (6). The credit economy in science emphasises the desire that scientists have to receive credit. Credit-seeking in science means prioritising one's job-security and career prospects when conducting research. The "credit" can take the form of the "prestige" that comes with being published in widely-read academic journals and being cited in other works. These publications and citations then signal their aptitude as a scientist to peers in their field. This perceived aptitude then increases the likelihood of that scientist gaining further grants and having access to higher-status jobs. As Huebner & Bright (2020) explain, these factors build on each other and stabilise over time, confirming the prestigious standing the scientist has within their scientific community.
Relatedly, researchers face decisions regarding the peer review system. Peer review in academia is a quality-assessment process. In seeking to publish a paper, a researcher first must find an appropriate academic journal. As well as the journal being of some prestige -- as we have seen --, it must publish papers of the pertinent field. Indeed, publications are made in increasingly specialised academic journals (ARL, [1998](https://www.arl.org/wp-content/uploads/1998/03/to-publish-and-perish-mar98.pdf)). The rewards system outlined above incentivises publications in academic journals that specialise in the field one wants to succeed in. We can see this in anecdotes Poteete et al. (2010: 405) provide, as well as empirical studies. For example, Siegel et al. ([2007](https://www.maxwell.syr.edu/moynihan/cqrm/qmmr/Table_of_Contents_5_1/)) find this is the case in political science. Furthermore, the peer review process is conducted by specialists who might view diversity of methods with scepticism (Lohmann, [2007](https://doi.org/10.5281/zenodo.997362)).
The above features, together, form what is commonly known as the culture of "publish or perish". This culture promotes the pursuit of credit, increased specialisation and heightened procudctivity, amongst other things. This culture is deemed pervasive in academia and is arguably problematic for scientific integrity (Fanelli, [2010](https://doi.org/10.1371/journal.pone.0010271); Anderson et al., [2013](https://doi.org/10.1007/978-94-007-5836-0_5); Heesen & Bright, [2020](https://doi.org/10.1093/bjps/axz029)). In the particular context of MEGs, the publish-or-perish culture raises a series of complications that I will introduce in the following section.
This section has shown that academic norms shape the activities of individual researchers. The publish-or-perish culture ultimately complicates the decisions that researchers make throughout their careers, particularly when operating within MEGs. It is necessary, then, that we account for academic norms when discussing the ontology of MEGs.
#### Against MEGs
In this section, I spell out how the publish-or-perish culture comes to play a critical role in our understanding of MEGs. I do so by discussing what Lee ([2020](https://doi.org/10.1086/710615)) calls *the reference class problem for credit valuation in science*. This will clarify practical aspects of the reward structures and peer review system discussed in the previous section. I will then relate this to the training-expertise trade-off discussed by Poteete et al. (2009). This section provides further substance to the claim that MEGs are conceptually maximalist.
A critical question that MEGs face relates with how different disciplines evaluate scientific outputs and give credit. This is what Lee ([2020](https://doi.org/10.1086/710615)) calls *the reference class problem for credit valuation in science*. The problem arises when a researcher belongs to different communities that disagree on how to evaluate research outputs. Lee focuses on scientific communities with "nested structures", where subdisciplines are nested within wider disciplines. The boundaries between these subdisciplines are much fuzzier than those we are speaking of in MEGs. Lee mentions, for example, the rise of solid state physics as a subdiscipline of applied physics (1029). The distinction between the two is much less clear than the boundaries we find between the computational sciences and classicism. With our paradigmatic case of the digital humanities in mind, we can see how the rewards structure described earlier on can be problematic. Indeed, if we consider researchers to prioritise credit-seeking and career stability, we find that scholars prefer purely disciplinary projects over multidisciplinary ones. The problem becomes more salient for those who actually join such projects. This is because, when the project is finalised, the research team will seek to publish their work. At this point, different opinions as to where the work is published will arise -- if they hadn't already. Both the team of classicists and the team of computational scientists, in our example, will have different preferences as to where the work is published and how findings are communicated. (A footnote could be added here discussing the nuances of the credit economy in the humanities and the rise of altmetrics).
The reference class problem for credit valuation in science is consistent with the trade-off that Poteete et al. (2009) identify in academic careers. The trade-off is between breadth of knowledge and specialist training. With the peer review process encouraging studies in narrow fields and a rewards system that promotes the publication of papers in such fields, it is advisable that researchers pursue more specialist training rather than multidisciplinary projects. Furthermore, researchers show reticence before activities that are critical to MEGs, such as collecting data from different sources, exploring new bodies of literature and employing diverse methodologies (Heesen, [2018](https://doi.org/10.5840/JPHIL20181151239); Poteete et al., 2009: 406). Such activities are problematic for at least two reasons. Firstly, they are extremely time-consuming and, therefore, counterproductive in an institution that rewards speed of publication. Secondly, and as mentioned earlier on, methods from other disciplines can result in scepticism on the part of deeply specialised peer reviewers.
Both the reference class problem and the breadth-specialisation trade-off are crucial factors in our study of MEGs.
### Towards a Planning Theory for MEGs
I have so far provided a brief overview of academic norms that entail a series of challenges for MEGs by shaping individual researchers' plans. The challenges identified have been referred to as *the reference class problem* and *the breadth-specialisation trade-off*. These are only two such challenges and needn't be considered exhaustive. Nonetheless, with MEGs working within academic environments and their members following academic norms that entail such challenges, I have argued that an account of the ontology of MEGs necessitates some account of the challenges they face.
Despite these challenges, we have seen that MEGs are increasingly popular, and major funders are only becoming more interested in multidisciplinary approaches to knowledge-production (Resnick, [2011](10.3389/fpsyt.2011.00020)). In what follows, I outline the "rationality norms" Bratman deems necessary for "modest sociality." This is to complement a Gilbertian account of MEGs that makes sense of their ontologically maximalist status but does not offer an account of the complex nature of the individuals within groups.
Bratman provides a framework for making sense of complex plans held by individuals and their "meshing" together to result in shared intentions. He suggests "the glue that binds team members together" (Cohen et al., [1997](https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.55.379): 2). The "social glue" (Bratman, [2014](https://doi.org/10.1093/acprof:oso/9780199897933.001.0001): 87) aims to tie members together when partaking in "modest sociality." This modest sociality has already been shown to be rather irrelevant in the case of MEGs, as I have argued for Gilbert's "ontological maximalism." Nonethless, the “norms” he describes as the “social glue” that make “modest sociality” possible are arguably useful for the present account. Indeed, we need something to account for the proliferation of MEGs in an academic environment that might be considered rather hostile. The four norms Bratman identifies are: agglomeration, consistency, coherence and stability. In this section, I will describe how each one is natural to our understanding of MEGs.
#### Social Agglomeration in MEGs
Bratman introduces *social agglomeration* as the agglomeration of individuals' intentions into a larger social plan.
It is worth reconstructing this statement following Bratman's own distinctions and example. Firstly, Bratman distinguishes "intentions" from "goals". Similarly to the plans and subplans described in section [**X above**], intentions relate with the much more immediate task at hand, whilst goals are broader and what can enter into conflict. Indeed, Bratman takes conflict to not hinder his conception of social groups (29). Bratman reintroduces the ecample of painting a house to make this distinction clear (*ibid.*).
When jointly painting a house with somebody else, Bratman explains, one's intentions relate with painting the house. Both participants in the joint action share this intention. However, the goals of either party needn't be consistent. One might want to paint the house to later sell it for a profit, whilst the other wishes to paint the house to later donate it to the historical conservation society. Bratman characterises these goals as non-intentional.
In the case of MEGs, this idea fits rather neatly. On the one hand, Bratman provides a framework that clearly accounts for the potential conflicts we described earlier on. On the other hand, Bratman provides a concept that describes the problems of MEGs: non-intentional goals. The way the concept is relevant to MEGs regards how individual researchers join projects with different goals in mind. One might join a MEG for different reasons depending on their career stage, for instance. For a principal investigator who leads a MEG, it may be an opportunity to impress their department and further their career within their own team. For an early-career researcher on their first postdoctoral contract, working in a MEG can give them exposure to different practices across different disciplines and help them decide where to specialise in next.
The reference class problem also maps onto the idea of non-intentional goals. Ultimately, the decision about where to publish a paper or scientific findings needn't detract from our successully jointly working in a MEG. To quote Bratman: "We might proceed with the joint house panting and leave to later a decision between the housing market and the conservation society" (29).
#### Social Consistency and Coherence in MEGs
Bratman holds social consistency to relate to two aspects: (i) intentions and (ii) "beliefs about matters of possibility and effectiveness" (30). When discussing agglomeration, we saw what intentions are and how inconsistent *goals* are not problematic for joint action. Indeed, Bratman often speaks of "agglomeration and consistency" as one entity. When it comes to "beliefs about matters of possibility and effectiveness", Bratman refers to background assumptions for the successful pursuit of some joint action. Bratman offers the example of disagreeing on what colour to paint a house as inconsistency in this regard. Bratman proposes deliberation and the eventual agreement on some "shared policy" [this needs a footnote] that allows the house-painters to proceed.
Bratman focuses mostly on cases where these background assumptions are indeed consistent. When introducing the concepts of "shared policy" and deliberation, Bratman juxtaposes "consistency and coherence". I will do the same.
In the context of MEGs, consistent and coherent "beliefs about matters of possibility and effectiveness" can be interpreted as shared methodological standards. Bratman's solution to account for social groups that are not "consistent and coherent" is to introduce "deliberations" and "shared policies" [this needs a footnote]. However, Bratman does not engage with these questions very seriously. ~~Much like this essay, ~~Bratman seeks an ontological account of social groups. The question of deliberation falls into the camp of social epistemology, [something we discussed early on in this paper]. Bratman mentions past works in which he expands on *deliberation* and insists that a group may conclude things that none of its members agree with ([1992](https://www.jstor.org/stable/2254116): 7; 2014: 147). This claim complicates Bratman's individualism, with *deliberations* seemingly filling a gap between members and their groups that is unproblematic when we consider the arguments for ontological maximalism.
Whilst social consistency and coherence are important to Bratman's *modest sociality*, they seem incompatible with epistemically diverse MEGs. However, these norms do turn our attention to the conceptual complexity brought about by "deliberations".
#### Social Stability in MEGs
Social stability refers to constance throughout time. Bratman sees it as following a sort of "snowball effect: once we are embarked on our shared activity, there will frequently be new reasons to continue" (90). These new reasons amount to the continued desire to successfully complete one's plans. These individualistic -- rather than social -- continued desires are what allow for stability. They also mean seeking that our sub-plans mesh and interrelate in an optimal way for our plans to be met. The snowball effect becomes social when we acknowledge this need for meshing sub-plans with others.
Epistemic diversity appears, once again, as a complicating factor when applying Bratman's norm to MEGs. This is because divergent sets of plans -- e.g.: those of computer scientists and historians -- do not seem compatible. One solution for this is *epistemic trust* as discussed [earlier on in section X]. Epistemic trust, we recall, is about trusting in the reliability of other scientists' work. The problem we found is that epistemic trust relies on shared methodological standards. We dismissed this problem by focusing on the academic context within which MEGs operate. We are now at a place where we can elaborate on this point.
Indeed, we can now refer to the publish-or-perish culture as providing stabilising norms. Epistemic trust is placed in one another not on the basis of some shared methodologies but on the basis that we expect our team mates to do work that is well-perceived by their own communities. The reference class problem is a way for researchers to allow colleagues of other fields to do their work without questioning it. This is further encouraged by their extremely specialised training, which does not in principle allow them to engage meaningfully with colleagues from other disciplines.
**With this, we have set the *desiderata* for MEGs' diverse plans to interweave in a productive way. In the final section, I propose a type of *epistemic humility* that is conducive to these *desiderata*.**
## Social Epistemic Humility
1. Define is epistemic humility (800 words)
2. Give an ordinary example (we disagree about who owes what in a bill, and my confidence drops because I am humble; 300 words) – this is it, but we can go a step further by suggesting a methodology we agree on to establish the truth)
3. Build up to a MEGs example: keep details vague but explain what the differences are with the ordinary example (my confidence drops because "my field makes mistakes", NOT ME")
4. We can go a step further by suggesting a methodology we agree on to establish the truth. The method to establish this methodology is beyond the scope of this paper.
We have so far established that MEGs can be conceived of as plural agents. This is in light of Wilholt's notion of *epistemic trust*: scientists place trust in scientific groups by virtue of their peers following widely accepted methodological standards. Dang's analysis also helps make this point: research groups ultimately make judgements that are not *prima facie* reducible to the judgements of their individual members. However, we have also seen some of the challenges that scientists in MEGs can face; namely, the reference class problem and the breadth-specialisation trade-off. In this section, I argue for a notion of *epistemic humility* that can help towards overcoming the aforementioned challenges.
**Defining epistemic humility**
Medina (2013) defined epistemic humility as "[involving]attitudes and habits conducive to the identification of cognitive lacunas" (43). He argues that epistemic humility is a virtue that improves knowledge-production processes by requiring that one qualifies their beliefs, identifies what they need to overcome cognitive limitations, and formulates questions to themselves and others. Consider the following example:
>Sam and Taylor just shared a meal at a restaurant. They each ate different menu items and are now splitting the bill. A quick check through the bill is enough for Sam to think they have to pay £17, and for Taylor think they owe £14. However, it is a £33 bill, so the numbers don't add up (17+14=31). Taylor then inspects the bill more closely and concludes Sam's meal cost £19, not £17.
In this scenario, we can narrow Sam's possible response to one of three:
* Sam profusely disagrees with Taylor's conclusion and maintains the view that they owe £17 without further examining the bill; or
* Sam wonders if they made a mistake and re-examines the bill more carefully; or
* Sam accepts Taylor's conclusion uncritically.
The first amounts to what Medina proposes is the epistemic "vice" that opposes epistemic humility: *epistemic arrogance*. Briefly, this is "the thought of being in full cognitive control os reality" (41). Epistemic arrogance does not leave space for others' views to be considered seriously.
The second option is where we see epistemic humility in action. It might be the case that Sam is aware of their own limitations in doing mathematical calculations in their head, or perhaps they trust Taylor a deal great enough to know to check the numbers again. Epistemic humility is present due to this self-evaluation and re-evaluation of the evidence present.
The third option exemplifies where Medina identifies epistemic humility as "pathological" because it "undermines one's confidence and erodes one's character" (43). This clarifies a negative definition of epistemic humility: it is not about blindly granting others cognitive superiority. Dormandy (2020) has called this being "intellectually servile."
In what follows, I apply this notion to MEGs.
**Epistemic humility and MEGs**
Note that the case of Sam and Taylor at the restaurant is far too simplistic to study epistemic humility as defined in the context of MEGs. However, the two challenges we identified for MEGs provide a framework for its application. Let us begin with the breadth-specialisation trade-off problem.
Recall Poteete et al.'s (2010) evidence-based argument for the great incentives scholars are given to specialise in unique methodologies. Conversely, "the great investment required to master any single method [...] limits the number of methods in which any individual can be expected to gain and maintain competency" (20). This is what motivates the study of MEGs in the first place: multidisciplinarity is required to tackle the big questions of our time, but their members do not share methodological standards. We had already identified that epistemic trust can be attributed to fellow MEGs whose methodologies we do not – as individuals – entirely grasp. The question now is: how do we promote that trust amongst members of MEGs who – similarly – do not share methodological standards?
Epistemic humility promotes trust amongst members of MEGs. The link between epistemic humility and epistemic trust is not novel. Consider Dormandy's ([2020](https://www.taylorfrancis.com/chapters/edit/10.4324/9781351107532-31/intellectual-humility-epistemic-trust-katherine-dormandy)) two-prongued argument. On the one hand, epistemic humility's function of identifying one's own cognitive lacunas can lead to seeking out the tools to fill such gaps. In other words, epistemic humility promotes self-improvement. This self-improvement leads to one's own greater epistemic trustworthiness. On the other hand, Dormandy argues that this same ability can inform individuals of their capacities to meet the expectations of those who invest epistemic trust in us.
We have already seen Dormandy's first link between epistemic humility and trust in action: Sam identified their own cognitive limitations and sought a method to fill the gap; Sam went through the bill a second time, this time more carefully. The second argument Dormandy proposes would be the case if either Sam or Taylor was the only person with access to the bill. The other party would need to trust them. The person with the bill would demonstrate epistemic humility by seeking out the tools that fill in the gaps for their own cognitive lacunas. They would do whatever is necessary to earn their counterpart's trust.
Dormandy's account is useful to the present task insofar that it promotes epistemic trust by drawing on epistemic humility. However, it is highly individualistic. We need an account of epistemic humility that is in accordance with our findings so far: that MEGs are not reducible to their individual members. With this, we can supplement the previous definition of epistemic humility with the following idea:
> What members of MEGs are humble about includes the intellectual lacunas of their domain of expertise.
With this, epistemic humility promotes uncovering the intellectual lacunas of one's own domain. Acknowledging the breadth-specialisation trade-off, one can then proceed to seek to fill in the gaps not by investing in years of training in the necessary area, but by seeking out the necessary knowledge in researchers from other domains. Thus, epistemic trust is promoted by being humble about what one's own field can achieve.
It is worth emphasising once again that epistemic humility not become "pathological." This is important in the academic context where perceived "brilliance" can often mean struggling with "impostor syndrome", especially amongst early-career and female scholars (Muradoglu et al., [2021](https://doi.org/10.1037/edu0000669). *Impostor syndrome* is present where "someone is talented, and has turned her talent into externally recognised success, yet reasonably doubts herself" (Hawley, [2019](https://doi.org/10.1093/arisup/akz003): §V). Although it is beyond the scope of this paper to extend this discussion any further, consider that directing humility at one's field can help reduce self-doubt, insofar that it is one's field – not oneself – what is being critically evaluated.
The case of epistemic humility to overcome the breadth-specialisation trade-off sets the necessary foundations for understanding its role before the reference class problem. Recall the four features of epistemic humility:
* Epistemic humility *is* about taking the views of others seriously;
* Epistemic humility *is* about re-examining one's own claims to identify and fill "cognitive lacunas;"
* Epistemic humility *is not* about being "intellectually servile"; and
* Epistemic humility *includes* critically evaluating one's own field of expertise.
[Basically, if everyone in academia demonstrated epistemic humility, MEGs could published and members of MEGs could be adequately credited because everyone would accept that there are some methodologies they just can't follow.]
### Objection: Too demanding
I have suggested that epistemic humility be held by individual academics to overcome institutional challenges to MEGs. This can be objected to on the grounds of being too demanding. Consider the specific argument for epistemic humility as a value to overcome the reference class problem. This problem is rooted in academic norms that are widely prevailant. I seem to be suggesting, thus, that all parties in the academic system – *i.a.*, junior and senior academics, publishers and funders – uphold the virtue of epistemic humility. This suggestion, the objection goes, is simply not feasible.
This is a useful objection that prompts the following qualification to my claim: epistemic humility is an epistemic virtue that consitutes but one possible solution to the reference class problem for MEGs. It is a necessary but not suffiecient condition for overcoming the reference class problem. Furthermore, epistemic humility has not been the end-game, but a means to achieve the epistemic trust that I claim is necessary for the success of MEGs. To this effect, we can qualify the need of epistemic humility amongst *only* those working within, or evaluating the work of, MEGs. Finally, the infeasibility of "everyone being epistemically humble" is an empirical claim. Whilst it might be difficult to imagine everybody in the academic ecosystem following such a virtue, it requires empirical observation to claim that this is simply not feasible. I might take this insight, though, to suggest that epistemic humility serve as a virtue that inspires future governance policies in academia. Given the limits of the present essay, I will leave that for another time.
### An Objection for Later On
The mention of trust as the end goal for the success of MEGs raises a further objection. I seem to have suggested that new-formed MEGs, when drawing on the work of other research groups, need not critically evaluate that work. They can just assume it was done right because they behave in the expected manner. The suggestion that MEGs trust one another by virtue of acting in a way that accords with academic institutions is problematic. It seems I suggested a "pathological" form of trust. There are two responses to this objection, each of which further clarify the role I grant to *epistemic trust*.
Firstly, consider the "efficiency argument":
1. MEGs must draw on great deals of research from different research groups
2. Research groups have more than one member
3. Therefore, there are more researchers than research groups
4. Therefore, it is more efficient to trust entire groups rather than each of their individual members.
This clarifies the motivation for my notion of *epistemic trust*. Epistemic trust as attributed to entire research groups facilitates the advancement of science by not needing to trust in the work of each of their members. For example, it is much more efficient to trust that CERN followed best practice when identifying the Higgs Boson in 2012, than seeking to trust each of its "5,154 autho" (Castelvecchi, [2015](https://doi.org/10.1038/nature.2015.17567)). Being "uncritical" about other research groups' work would only support this desire for efficiency in MEGs.
Secondly, I did not explicitly call for an "uncritical" trust. Medina's framework is helpful here. To avoid some "pathological" epistemic trust, effort must be spent in attributing that trust. Recall Sam uncritically accepting that they owed £2 more than they'd initially thought. This was precisely *not* what epistemic humility is about. Conversely, epistemic trust is *not* about trusting other research groups simply because they followed some set of shared academic norms. Indeed, Wilholt implies there is effort in this process when saying one "*invest[s]* epistemic trust" ([2013](https://www.jstor.org/stable/24563046): 233; italics my own). It is not a mere deposit or automatic attribution. Trust must be earned by research groups. The prior discussion about shared norms is only the first step. What the next step looks like is beyond the scope of this paper.
# Word-Count
:::info
As of `22/02/2022 10:00`: **609**
As of `22/02/2022 12:18`: **3,848**
As of `22/02/2022 12:48`: **6,139**
As of `24/02/2022 10:59`: **6,351**
As of `24/02/2022 13:17`: **6,397**
As of `14/03/2022 10:39`: **6,544**
As of `14/03/2022 13:05`: **6,717**
As of `01/04/2022 16:12`: **6,996**
As of `01/04/2022 16:35`: **6,865**
As of `07/05/2022 18:54`: **8,683**
:::
:::danger
Limit: **10,000**
:::