The Mutual Knowledge Problem and Its Consequences for Dispersed Collaboration

Catherine D. Cramton, The Mutual Knowledge Problem and Its Consequences for Dispersed Collaboration, Organization Science 12 (2001-06). (pdf) (notes)


This paper was very interesting, especially since I just spent the year working on a team that was split mostly into two locations (St. Louis vs. Oklahoma City), with a few other players in other cities around the country. That's why I sought it out after seeing it as a cited paper in N. Sharon Hill and Kathryn M. Bartol, Five Ways to Improve Communication in Virtual Teams, MIT Sloan Management Review (Fall 2018) (notes). It's difficult to work on distributed teams and get it right, so I wanted some advice.

I've grabbed lots of notes, quotes, and references here.

Five problems to achieving a state of mutual knowledge were identified:

  1. failure to communicate and retain contextual information
  2. unevenly distributed information
  3. difficulty communicating and understanding the salience of information
  4. differences in speed of access to information
  5. difficulty interpreting the meaning of silence

(2) and (4) are obvious enough. Skip.

Point (5)—misunderstanding silence on the other end of the line—was the most surprising, and perhaps the most interesting. I made some notes about it here: Interpreting silence. It's really obvious—once you're aware of it. Think of any time you've been ghosted for any reason, deserved or not. It hurts. I've been trying to be conscious about it.

An interesting point was raised about (1), the "failure to communicate and retain contextual information". A reference was made to information sampling (Garold Stasser and William Titus, Pooling of unshared information in group decision making: Biased information sampling during discussion, Journal of Personality and Social Psychology 48:6 (1985)). The quick idea is this: when people want to talk about something, they sample from the information that they know. If more people in a group have the same piece of information, it is more likely that it will be mentioned, which leads to more people knowing the shared information. Esoteric information is less likely to be shared, and fewer people will know it. If a team is split up into separate parts, information is less likely to be sampled and shared across partitions because much of the communication happens locally, so different teams will be holding a different pool of knowledge—and that means that they'll be working in different contexts as they try to solve the group's tasks.

So you need a Single Source of Truth even more with a distributed team. Imagine what kind of abuse I get when I suggest that teams should have an internal blog. But they should. Individual knowledge and team knowledge need to be mixed together.

Finally, point (3): misunderstanding which information is important. Similar to (5), this is one I've clearly been getting wrong. An example given in the paper is how easy it is to misunderstand things written in emails. A sender might think one point given is obviously the most important thing; the receiver might think something else is the most important thing and not even see the important thing the sender wanted to communicate. The point needs to be clearly The Point, else frustration and bad decisions will soon follow. It's much worse when you're separated from the person you're trying to communicate with because all of the hand gestures and facial expressions and voice tone and so on that are also communicating information aren't there.

One more final note from the paper that raised a point that I often get wrong: attribution of blame. I get that it's not helpful to blame people directly, instead of the situation or result, but I make that slip often. There's even a field of study related to it—Attribution theory—so I guess I'm not the only offender. Anyway, from the paper:

When an error or conflict in information exchange is detected, people make attributions concerning its cause. [...] Personal attributions associate the cause of the communication conflict with some characteristic or behavior of an individual. [...] Attributions were judged to be constructive if they facilitated inquiry and change to reduce the incidence of communication conflicts in the future. Attributions were nonconstructive if they were task irrelevant or destructive to cooperation, inquiry, and adaptation. The researchers suggested that situational as opposed to personal attributions tend to produce better resolution of conflicts because they focus participants on modifying the "contracts that guide the communication process. If attributions are destructive, contracts concerning the communication process break down and people withdraw from cooperation.

If you work in distributed teams you should read the paper. It might keep your foot out of some of the traps that others have stepped in. (Now to find ways to pry my foot out of these traps that I just noticed...)


Aside. Interesting-looking rabbit hole discovered but not pursued: Several references are made to Herbert Clark, a psycholinguist at Stanford. Based on titles alone, I would check out the following if they ever line up with what I'm pursuing:

Leave a Reply

Your email address will not be published.