People say "I think" a lot. Here are some examples:

• I think you brought me the wrong order.
• I think the numbers in the report are wrong.
• I think you need to turn left at the light.
• I think we need to replace the whole water heater.
• I think iPhones are better than Android phones.
• I think you should quit your job and start a business.
• I think that kids shouldn't have any screen time before the age of five.
• I think everyone should take vitamin D supplements.

I don't think that it is always problematic to use that phrase. In fact, I think that it is often appropriate. However, I also think that most people would benefit from tabooing the phrase "I think" in various situations.

Actually no, let me rephrase that: I am moderately confident that a large majority of people who engage in somewhat intellectually serious discussion would benefit from tabooing the word "think" a moderate amount more frequently than they currently do.

Accuracy

One potential issue with saying "I think" is that it's just not clear what you mean. Someone saying "I think the numbers in the report are wrong" might be saying that they're 99.9% confident that the numbers are wrong. On the other hand, someone saying that they think you should quit your job to start a business might only be 20% confident that you should in fact go down that path.

On first approximation, something seems really broken here. 99.9% and 20% are very different levels of confidence. Why would we allow a phrase ("I think") to point to such wildly different underlying concepts ("99.9% confident" vs "20% confident")?

Well, here's the thing: context can be very powerful. Like when a diner says to a server, "I think you brought me the wrong order", in that context the server is not going to wonder to themself: "Hm, is this person saying that they are 30% confident that the order is wrong? 75%? 99.99?". No. They are going to correctly assume that the diner is communicating that they are almost positive that the order is wrong.

With that said, I don't expect that the phrase "I think" will usually lead to meaningful miscommunications about accuracy. I think that humans have mostly figured it out and that in most situations, context is enough to allow for effective communication.

But that doesn't mean that there isn't room for improvement. I estimate that there are in fact many situations where using a phrase like "I think" leads to meaningful miscommunications about accuracy.

For example, if Alice tells Bob that she thinks he should quit his job and start a business, I could imagine Bob taking it as "Alice is being serious here and is probably 90%+ confident in making this recommendation" when in reality Alice actually meant something more like "I'm just thinking out loud and get a vague sense that doing this will be better for you, but there are a lot of factors that determine whether it actually is better for you and I haven't spent nearly enough time considering them all to be more than 30% confident in this recommendation".

I can't think of a good way to make a real argument here though. I'm not even sure what the specific claim is that I want to make. It's something along the lines of what I said earlier about how the phrase "I think" leads to meaningful miscommunications about accuracy. It's hard to say what I mean by "meaningful" though. It's hard to say how often I expect this sort of thing to happen, and how much damage I expect it to cause.

And even if I can zero in on a specific claim, I'm not sure how I would make an argument for it. Make a copy of my life experience and upload it into your brain? No. I guess a way to approximate that would be to come up a bunch of good examples that serve as effective intuition pumps, or to link to 250 concrete examples of this plausibly happening in eg. LessWrong discussions, but I'm struggling to come up with such examples.

So with all of that said, I think the statement I want to make to you, the reader, is that you should consider tabooing the phrase "I think".

Identity

Paul Graham wrote an essay called Keep Your Identity Small. The idea is that, in my own words, people become dumb when they talk about things that are a part of their identity.

Think about a diehard Cowboys fan arguing that their team will win the Super Bowl, a socialist arguing about tax policy, or one of those people from JW.org standing on the street arguing about God.

Using terminology from Julia Galef's book The Scout Mindset, because their identity is tied up in these things, they act like soldiers instead of scouts, fighting for their current beliefs instead of being open to go wherever the best arguments take them.

Alternatively, using terminology from Tim Urban's book What's Our Problem, when beliefs become part of your identity it pushes you further down the vertical axis, leading to you approaching things less like a scientist and more like a zealot.

I think we can agree that we want to be like scientists, not zealots. Scouts, not soldiers. That we want to be to open to wherever Bayes' Theorem takes us and not let our identities get in the way.

In pursuit of these ends, I think we need to keep a somewhat close eye on our use of language. I have the post Rationality and the English Language in mind here:

If you really want an artist’s perspective on rationality, then read Orwell; he is mandatory reading for rationalists as well as authors. Orwell was not a scientist, but a writer; his tools were not numbers, but words; his adversary was not Nature, but human evil. If you wish to imprison people for years without trial, you must think of some other way to say it than “I’m going to imprison Mr. Jennings for years without trial.” You must muddy the listener’s thinking, prevent clear images from outraging conscience. You say, “Unreliable elements were subjected to an alternative justice process.”

...

My point is not to say that journal articles should be written like novels, but that a rationalist should become consciously aware of the experiences which words create. A rationalist must understand the mind and how to operate it. That includes the stream of consciousness, the part of yourself that unfolds in language. A rationalist must become consciously aware of the actual, experiential impact of phrases, beyond their mere propositional semantics.²

Circling back to the phrase "I think", I worry that this phrase not infrequently muddies our thinking. More specifically, I worry that it happens when we say "I think" in order to communicate a low confidence belief.

Let's take "I think everyone should take vitamin D supplements". This is actually a belief that I hold. Well, "everyone" is probably too strong, but let's roll with it.

This belief of mine is a relatively low confidence one. Back when Covid was more prevalent and I would read about stuff and I kinda sorta remember reading that vitamin D is something that 1) many people are deficient in, 2) such deficiencies are actually problematic, and 3) supplements don't really have any (physical) downsides. But my memory is fuzzy, those assumptions aren't ones I spent much time investigating, and I haven't thought through what other factors might be worth considering, nor how much weight those factors have. Nor have I battle tested my beliefs in discussions with smart people, nor have I looked into what the recommendations of (properly incentivized) experts are.

But if I go around telling all my friends "I think everyone should take vitamin D supplements"... I dunno... I worry that I'd become a little bit attached to that belief. Subtly, and unconsciously. I don't want it to be the case, but I suspect that it would be. Both for myself and for humans in general.

I actually make somewhat of a conscious effort to "be the type of person who changes their mind". Like, I'll go out of my way to announce it when I change my mind on something, or just when I update my beliefs at all. I also try to identify as someone who constantly is updating incrementally. So yeah, I make a little bit of an effort to announce these sorts of things.

But I'm not perfect. I'm an aspiring rationalist, as they say.

So, with all of that said, I think it probably makes sense -- for myself and most others -- to lean away from using phrases like "I think" when they risk communicating more confidence than you actually have.

But I'll give the same hedge that I gave in the previous section. I'm having trouble formulating a specific claim here, and I'm having trouble making a real argument. And so I guess what I'm looking to do in this post is kinda just to bring up the topic, vaguely gesture at a claim, describe some feelings about that claim that hopefully serve as a little bit of an intuition pump, see if those feelings resonate with the reader, and propose that the reader consider tabooing the phrase "I think" more frequently.

Alternatives

Ok, so suppose that you are convinced. Suppose that you buy what I'm selling and want to start tabooing the phrase "I think" more often. How can you do that? Well, I'm not totally sure, but I'll take a stab at it.

To start, I think that it is important to distinguish between statements of belief and statements of value expression. I have Robin Hanson's futarchy in mind here, where the slogan is "vote on values, but bet on beliefs". Since hearing about this idea, the distinction between beliefs and values has always really stood out to me.

A statement about a value is something like "I think that scientific progress is valuable in and of itself". On the other hand, a statement about a belief is something like "I think that spending on medical research improves health outcomes more than spending on preventative care does". In other words, statements of beliefs are predictions.

I don't really see anything problematic about using the phrase "I think" when making statements about value. I guess to be more clear you can say something like "I personally value scientific progress" because the initial statement might mean that you personally value scientific progress, or it might mean that you see it as something that has inherent value.

For statements of belief, I see two approaches for replacing the phrase "I think": quantitative and qualitative.

To take the quantitative approach, you can assign a probability. Like, you can say "I think X is 90% likely" or "I'm 90% confident in X". You can also be a little handwavvy and say that you're "something like 70-90% confident in X".

But putting a num on it can be weirdly difficult. I feel like it shouldn't be, but I know that I sometimes just can't bring myself to do it. Sometimes I just keep flip flopping ("90%. No, 70%. No, 80%. No, 65%."). Other times I just can't even bring myself to come up with an initial estimate. At times like these, taking the qualitative approach is a huge help.

You can say that you're "pretty confident". You can say that you're "somewhat confident". That you "suspect X". That you "wouldn't be surprised by X". That "X seems plausible". That you "think X is overwhelmingly likely". I'm sure there are a bunch of other good adjectives to throw around.