A short primer on discussions

Good discussions can improve our understanding of the world, clarify our values, and develop the concensus required for taking action as a group, whereas bad discussions tend to aggravate, alienate, and confuse. Great products, businesses, and policies often rest on a foundation laid in good discussions.11A debate is a different beast. A debate is for the benefit of an audience and often something you try to win in the eyes of that audience, regardless of what is true.

Though I haven’t found a fool-proof recipe for good discussions, I’ve noticed three basic mistakes that spoil them:

  1. Using ill-defined words.
  2. Accepting arbitrary beliefs.
  3. Confounding truth and preference.

Avoiding these mistakes is not rocket science, but it does require patience, precision, and discipline. They are particularly damanging because they strike at the core elements of a discussion: language, belief, truth, and preference.

Discussing the War in a Paris Café.Discussing the War in a Paris Café by Fred Barnard appeared in Illustrated London News on 17 September 1870. Wikipedia
Discussing the War in a Paris Café.22Discussing the War in a Paris Café by Fred Barnard appeared in Illustrated London News on 17 September 1870. Wikipedia


A discussion is futile if participants use the same word to mean different things. No common understanding can develop, which means progress will stall.

For example, “liberal” is commonly used both to refer to a proponent of a big, high-tax government and a proponent of a small, low-tax government. One reason for the confusion stems from the distinction between economic liberty (low taxes, free markets) and social liberty (unconstrained lifestyle). The political left is traditionally considered socially liberal and economically regulated (big, high-tax government), whereas the political right is traditionally considered socially regulated and economically liberal (small, low-tax, government). (This is a simplification however.) Thus, in a discussion about politics we need to be precise about what we mean by “liberal.”33Another reason why liberal is confusing is the Liberal Democratic Party in the United States. Here, “Liberal” with a capital “L” takes on the meaning of a proponent of whatever the party advocates, which might not be at all liberal the original sense.

There isn’t really a commonly accepted word for those who are both economically and socially liberal (i.e. not traditional left/right). They are sometimes called “classically liberal” and though some might refer to them as “libertarian” the latter of has stronger connotations with ultra-minimal states.

Defining terms takes time, but it pays off and you don’t have to do it all up front. You can ask for clarification in the course of a discussion: What do you mean when you say “simple colours”? As a guideline, try to not redefine words unnecessarily. Check dictionary definitions and other authoritative sources. You might want to give a word a more precise meaning than is common, but avoid contradicting authoritative usage if you can.


The second mistake is to think that all beliefs are created equal. Belief has a precise meaning here: one’s degree of belief in a statement refers to one’s assessment of the probability that the statement is true. Our degree of belief in something lies between 0 (certainly false) and 1 (certainly true). We form our beliefs based on our prior beliefs and the evidence we observe. We say that we are rational44It is more or less impossible to have a productive discussion with someone who does not strive to be rational (at least about the topic at hand). An irrational person can literally hold any arbitrary statement for true. The only discussion you should engage them in is one about rationality, but don’t expect much.

Rational people can still disagree about truth if they have observed different evidence, but I suspect this is the cause of fewer disagreements than irrationality or unwillingness to consider other people’s input.

when we strive to form consistent beliefs.

For example, most adults have a degree of belief in the existence of Santa Claus that is close to 0 and a degree of belief in the earth being round(ish) that is close to 1. If, however, we were to observe a sled with a white-bearded man in the sky, pulled by flying reindeer we would increase our degree of belief in Santa Claus (unless we had a good alternative explanation for the observation).

Beliefs lie at the heart of discussions about what is true. If people arbitrarily assign belief, there is no way for them to agree on what to believe given some observation. If, on the other hand, everyone strives to form consistent beliefs, then the group can incorporate each other’s experience, observations, and reasoning to form beliefs that are more accurate than those of any individual participant.

How to form consistent beliefs is formalised in Bayesian probability theory – a fascinating topic and the foundation of scientific reasoning – but you don’t need a deep grasp of it. An understanding that beliefs should be based on evidence and consistent reasoning will suffice for most discussions. In fact, many scientists do great work despite being unfamiliar with formal Bayesian inference. (See Further Reading for more details.)

Truth and Preference

There are actually two kinds of discussions:

  1. Discussions about what is true (how the world behaves).55Most people (including many, if not most, philosophers) would agree that there is an objective truth. If you don’t accept that there is such a thing as truth or rationality, I think you’ll struggle to have productive discussions with other people.

  2. Discussions about what is preferred (what is good and bad).

Discussions about what is true try to answer questions such as: Will a minimum wage lead to higher unemployment? and Are carbs unhealthy? Such discussions are about the evidence participants have observed and how they reason about those observations. The result of such a discussion is that each participant updates their beliefs, taking new input into account.

Discussions about what is preferred explore questions such as: Should same-sex marriage be legal? and Should we eat pizza tonight? Since preferences guide actions, if an individual has inconsistent preferences, they risk taking action that negate itself (or worse). Similarly, if a group is to take action together, they need to agree on the preferred outcome. Discussions can help with both problems. A discussion can investigate if a set of preferences is internally consistent. A discussion can also attempt to find a compromise; that is, a set of preferences acceptable to all individuals in a group.

A set of preferences can be inconsistent either by having logical contradictions (“abortion should be legal and all fetuses have an absolute right to life”) or by having practical contradictions (“abortion should be illegal and no unwanted babies should be delivered”).

Practical condradictions are particularly difficult to identify because they also rely on our beliefs about what is true. That means we need to intersperse discussions about truth (how the world behaves) with discussions about preference (what is good and bad). It is all too easy (and common) to instead confound the two by stating a preference for a course of action that you believe will result in the outcome you actually want.

For example, if someone simply states a preference for invading another country, you don’t have much room for discussion. If, on the other hand, they say they prefer the population of a country to live under a rule of law instead of a dictator and they believe the best method to achieve that is a military invasion, you can discuss both their preference and the beliefs that conclude that an invasion is the best way to achieve the desired outcome.66Some preferences can be impossible to satsify because of real-world constraints.

Therefore, we must clearly distinguish between truth and preference in order to have productive discussions. Doing this well takes discipline and practice, but it can mean the crucial difference between conflict and cooperation.


Good discussions challenge our beliefs and preferences, but in return improve the accuracy of our beliefs, identify inconsistencies in our preferences, and find compromises that allow us to take action together.

By using well defined words, by forming rational beliefs, and by clearly distinguishing between our preferences and our beliefs, we increase our chances of having good, productive discussions.

I sometimes think of discussions as a form of exploration in the world of ideas which, just like the exploration of our physical world, can be hard and uncomfortable, but also provide the excitement and joy of new discoveries.77The collaborative aspect of discussions means that the more you discuss with someone, the more understanding you get for each other’s language, beliefs, and preferences. That means you can cover ground faster than if you frequently had to stop to define terms or understand each other’s beliefs. That’s why discussions with long-time friends or colleagues can be particularly satisfying.

So, keep an open mind, be patient, and admit shortcomings readily.

Bon voyage!

Thanks to Jan Sramek and Emily Rookwood for suggesting improvements to drafts.

Further Reading