Intellectual Schelling Points


Part 4 of the essay series Research and Knowledge Accumulation


Towards accumulation

Left to itself, knowledge naturally decays. Yet we see many cases around us where researchers make discoveries and knowledge accumulates. What allows knowledge to accumulate in these cases?

In this essay, we will introduce the idea of an intellectual Schelling point. This will help us to understand the phenomenon of knowledge accumulation and more about when knowledge does and does not accumulate.

Intellectual common ground

For knowledge to accumulate among researchers, those researchers must have enough common ground. When you have enough intellectual common ground, you can share ideas, correct errors, challenge assumptions, and learn from each other. When you don’t have enough common ground, this is much harder.

Suppose for instance that you’re studying the history of the Peloponnesian wars and Greek trade routes, and I’m studying RNA replication. I might love hearing about your ideas and theories and you might love hearing about my experiments. But it would be hard for either of us to contribute much to each other’s research.

There is obviously a spectrum here. As common ground increases, you share more, can correct each other more, can get deeper into topics. You even remember more. As common ground decreases, there is less to share, less gets through on either side, it’s harder to identify the true points of disagreement, harder to correct each other’s errors.

The intellectual exchange that is enabled by intellectual common ground is useful for three distinct purposes, all of which relate to accumulation.

First, there is a function of generativity. When you have multiple people able to communicate with each other on the same topics, it becomes much easier to generate new ideas. These new ideas are the ore from which new discoveries can be extracted.

Second, there is an error-checking function. When multiple people communicate with each other on the same topics, they frequently become able to provide useful feedback, spot each other’s errors, question each other’s assumptions, and so forth. Error-checking is the refinement process that helps to extract what is valuable from the ore.

Third, there is a transfer function. When people communicate, they pass ideas back and forth. Good ideas checked for errors are the sort of things that need to be transferred for knowledge to accumulate.

A critical mass of interest and intellectual common ground are necessary conditions for knowledge to accumulate. Without interest, there won’t be anyone there to transfer the knowledge to. Without intellectual common ground, the transfer will not occur.

Seeking common ground

Finding intellectual common ground is sometimes easier said than done. Consider the example of happiness. Imagine that we’re interested in happiness and want to have a group of researchers investigate it. What trouble will we encounter?

To start off with, different researchers will have different conceptions of happiness. Some will think about happiness as a feeling. Some will think about it as a complex set of feelings and behaviors. Some will think in terms of a placeholder, perhaps “whatever it is that people are referring to when they use the word ‘happiness.’ ” Some may propose that “happiness” is a folk concept with no objective correlate.

Next, different researchers will favor different methods of investigation. Some will want to introspect and make phenomenological distinctions. Some will want to administer surveys to local populations. Some will want to make conceptual distinctions and try to create theories that capture our linguistic practices. Others will want to conduct experiments to try to discern the conditions under which happiness (as they conceive it) occurs.

If we concretely imagine a group of researchers like this trying to communicate with each other, it’s not that they won’t learn anything. They have some intellectual common ground. But it’s much less common ground than the single word “happiness” suggests. Compared to groups that are studying, e.g., patterns in bitcoin prices, computer vision for self-driving cars, or the spread of a virus, it is much harder for the happiness researchers to transmit ideas to one another, correct each other’s errors, and so forth.

Engagement prior to agreement

If having a single word as a banner is not enough to give researchers sufficient intellectual common ground, what does? Here there is a puzzle. The intellectual common ground that enables researchers to share ideas and check each other’s work is what produces agreement. It cannot therefore presuppose that agreement. Whatever it is that gives researchers the initial common ground they need must do so even if the researchers disagree.

There are in fact a variety of things that produce intellectual common ground even without presupposing agreement. Here are a few examples:

  • Observations — Researchers can agree that observations have occurred, or allegedly have occurred, without agreeing on the true interpretation of those observations. Those observations, or alleged observations, can then form the basis for successful intellectual exchange.
  • Methods — Researchers can use the same or similar methods, even while aware of the imperfection of those methods and disagreeing with each other about exactly why the methods are imperfect. Some methods can be used by different researchers in a way that yields similar enough results to create intellectual common ground.
  • Instruments — Researchers can use the same instruments and compare and discuss results even while disagreeing about how the instruments work or what the instruments are measuring. Some instruments will yield similar and interesting enough results to enable researcher discussion, communication, etc.
  • Theories — Researchers can discuss or act on the basis of a theory, even while all believing the theory is ultimately mistaken and while disagreeing with each other about how. Some theories will enable sufficiently fruitful discussion or yield sufficiently similar results when acted on to generate intellectual common ground.

This list is not meant to be exhaustive. It should also be clear that while some observations, methods, instruments, and theories generate intellectual common ground, not all do and not all generate the same amount. Many factors are relevant. How accessible are the observations? How easy are the methods to use? To what degree do the instruments yield the same results for different researchers? How easy is it for the researchers to understand the theories in the same ways?

Our happiness researchers might benefit from any of these things. If we had a set of tightly reproducible observations that seemed relevant enough to happiness, the researchers could set aside many of the other questions and jointly investigate the observations. If we had methods that enough researchers could use that generated observations that were close enough to one another, the researchers could focus their efforts on those methods and those observations. If we had theories that were easy to understand and easy to apply, the happiness researchers could use those as their jumping off points.

That wouldn’t be the end of their investigation. That would be the beginning, the point after which there was enough common ground for the researchers to talk easily enough, share more ideas, correct each other’s errors, and so on. Then, if the ideas were good enough and there was enough of the right sort of error-checking, knowledge would be produced and shared.

Intellectual Schelling points

It would be good to have a name for things that generate intellectual common ground even without people needing to agree. I propose that we call them intellectual Schelling points.

The idea of a Schelling point comes from game theory. A Schelling point is an answer that people tend to choose when they can’t communicate with each other. Imagine that we were we were presented with this diagram and asked to choose a square, knowing that we would win if we all chose the same square:

Image for post

If we weren’t allowed to communicate, which square would we select? The red square. It’s a Schelling point, a solution we’ll tend to choose absent communication. Or imagine that we were asked to select a specific meeting location in France, knowing that we would win if we all chose the same meeting location. If we weren’t allowed to communicate, what location would we select? Perhaps the Eiffel Tower. That’s another Schelling point.

Now if a Schelling point is a solution people will tend to choose absent communication, an intellectual Schelling point is something that attracts researchers and focuses them on the truth even absent agreement.

Intellectual Schelling points generate intellectual common ground. In the presence of such points, researchers come to be able to communicate with each other, share ideas, correct errors, and so forth. In the absence of such points, intellectual common ground is much harder to come by. Communication is less valuable, feedback is less valuable. Less error-checking occurs, fewer ideas are transmitted.

A crucial role

Given their crucial role in enabling discovery, error-checking, and communication, it may come as no surprise that intellectual Schelling points have played a crucial role in the history of thought. In the next essay, we will look at three famous examples from history to see the presence and absence of intellectual Schelling points at work. Then we’ll finally get to the hypothesis about what is going wrong with academic and scientific research.


Next essay: The Endurance of Research Programs
Previous essay: Knowledge Preservation and Decay