← back to essays | home



This essay does not advocate a specific metascientific point of view, and as such does not make any particular suggestions. It is, rather, a propaedeutic to any metascientific analysis that might attempt to make suggestions; it attempts to outline what the right questions to ask even are, by summarizing the various aspects of metascience and the ways in which these aspects interact.

That being said, I do take a specific perspective, viewing science as being done primarily by systems composed of institutions composed of individuals; the extent to which science thrives is largely based on the extent to which these institutions and the individuals within them align, the individuals working with their institutions rather than for themselves, the institutions working with one another rather than for themselves, and the systems working towards the goal of science rather than self-preservation. Based on this perspective, I largely view the metascientific question — how do we improve the way science is done? — as a question of the proper alignment of incentives.

This essay is a counterpart to the essay Metacognitive Science, which outlines a new program for improving scientific cognition, rather than scientific institutions; the two have a necessary interplay, which is detailed there.


I examine the extent to which institutions can (a) be successful, and (b) meet their ideals, by introducing two notions of alignment: internal alignment, or the extent to which the members of an institution are incentivized to take actions that lead to the success of that institution, and external alignment, or the extent to which an institution is incentivized to use the resources gained via success in pursuit of its ideals. To support the assumptions behind this view, I argue that success is often unrelated to the pursuit of ideals (so that internal and external alignment are indeed distinct), and that individuals and institutions can be expected to act selfishly for their own success (so that internal and external alignment, respectively, are indeed concerns).

I then examine the case of science, dissecting it into distinct institutional arrangements and norms. The doing of science is separated into three stages, and the cases of internal and external alignment among various individuals and institutions are examined at each stage. Then, as an example, I show how the reproducibility crisis arises as a particular failure of external alignment at a particular stage of the doing of science.

Institutional Success in General

For an institution to achieve its ideals, such as the creation of knowledge, it needs to be able to achieve success — to marshal resources at all — and to use these resources in pursuit of its ideals. It needs to be successful and externally aligned.

Internal Alignment

Of success and ideals, success is generally pursued with more vigor, but is pursued on multiple different levels: not only by institutions, but by people acting individually within them. People and institutions obey their incentives in the pursuit of resources and status — call the acquisition of these material success. (Though status is obviously itself a resource, people do not pursue it as a resource — purely instrumentally — but in itself). Of course, people tend to care more about their own resources and status than those of the institutions they are members of, but, as those institutions are their means for acquiring them, they are often incentivized to guide the institution down its own incentive paths.

The "often" case is generally when they have some sort of skin in the game — when the material success of the institution is directly tied to their own success, or when they are strictly managed by people who are strictly managed by .... by people whose success is tied to the material success of the institution. When this is not the case, they may do their own thing, or act in contrast to the short or long-term material success of the institution when it ensures their own material success.

This is to say that the material success of an institution relies on its being able to tie its success to that of its members, either directly through e.g. commissions, stock, or using their connection to it for status/connections, or indirectly through chains of management. Call this internal alignment. When we have multiple institutions acting in tandem, or when we can consider one institution as consisting of multiple independent parts, the issue of internal alignment applies on a second, intermediate level, and the success of the institution or system of institutions depends on its being able to coordinate these parts so that they do not act for their own material success to the detriment of the whole.

As regards the psychology of all this — why can people, even ideologically committed ones, generally be expected to act selfishly? — there are multiple reasons.

  1. Ideological commitment is usually an epiphenomenon: it's only really there to satisfy the demands of the ego/identity, which manifest it, and does not arise from a wish to see the world changed in a certain way. If the world were changed so as to completely satisfy the demands of the ideology, the person with the ideological commitment would simply make further demands; if a perfect, easy solution were proposed, they would reject it. (For instance, suppose that climate change was magically and permanently solved tomorrow. Most people who are worried about it would be relieved, but the people who get their salaries and status from it, whether they singled it out as a unique, apocalyptic threat or not, would not be). Given that it's an epiphenomenon, it's reasonable to believe that the reasons people consciously believe they have for doing certain things are not generally the reasons that actually motivate them, in the sense that if they were taken away they would stop doing the thing. Really, they are motivated by material success, likely for evolutionary reasons. (I think this is what Robin Hanson's constantly saying, but who knows).
  2. People focus on their specific part of the institution, seeing it as important to the exclusion of others. Hence, they can reason (whether as a mask for furthering their own material success or not), what they are doing is absolutely necessary to the whole venture, and resources and status ought to be funneled to it. Convenient!

Of course, there are some people — probably better characterized as individuals — who, whether due to an epiphenomenal ideological commitment that is nevertheless held tightly to the actual goals by a particularly self-critical ego, or a "genuine", non-epiphenomenal ideological commitment, do use material success instrumentally in order to achieve their ideals. (This kind of person would take a perfect, easy solution, though they may still not be satisfied). Call them true believers; Malcolm X is a great example.