In real life (certainly in this country, though not nearly as much in Australia), I've sometimes been accused of having no filter on what I say. This isn't true, of course, but the extent of my sociological observations goes farther than most people here. America is a country where it is crucially important not to notice things, as Steve Sailer put it. If you notice, you absolutely shouldn't comment. If you comment, you really truly ruly shouldn't dare find any of it funny or ironic, or indeed anything other than deadly serious.
But these are serious times, and joking with the world at large about the wrong things does not tend to get rewarded. One must pick one's audience, so to speak. This blog, for instance, is not that audience. Everything said here is said to everyone, for all time, and able to be quoted out of context and misconstrued for years to come.
But it is oppressive to never speak one's mind freely. Paul Graham recommended drawing a wall between one's thoughts and one's speech, the former being free, the latter being restricted for what is acceptable.
I dance a finer line. With people whose character I feel I can trust, I'll say what I think. Sometimes they're surprised, because this assessment isn't actually that correlated with how long I've known a person. Some people I know and consider dear friends never fall into this category. Some people I've known I a day or two do. Those, I think, are the ones who sometimes think I have no filter.
So what determines whether I think it's likely to be worthwhile to speak freely to someone or not?
As far as I can tell, there are three main classes of requirement.
The first is that you know, without me needing to explain it to you, in a deep and instinctive sense, the difference between the following words:
Is Correlated with
The statement 'all Australians are obnoxious' is very different from 'the average Australian is obnoxious'. People that don't get this will transform the latter into the former, and thus read it as 'he is accusing me of being obnoxious because I am Australian'. Conversation with people who think like this is always a minefield, so it's better to stick to small-talk.
Related to the above, understanding basic causal inference is equally important. Umbrellas are correlated with traffic accidents but do not cause traffic accidents - rain causes both. Prisons affect crime and crime affects prisons - prisons fill up when crime increases, and the increase in prison populations reduces crime.
You don't need to use words like 'omitted variables' and 'simultaneity', but you do need to have a good feel for these different types of models of the world, and be able to think about how they might apply to some new situation.
These requirements mean that your words aren't apt to be misconstrued. If you happen to get lazy and utter something like 'Australians are obnoxious' rather than specifying a precise probabilistic and causal statement, the person will not immediately assume the most inflammatory possible interpretation.
The second requirement is that you consider truth a near-complete defense to any charges levelled against pure statements about the nature of the world (as opposed to statements of opinion). If the average Australian is indeed obnoxious, one should be free to say so. You do not change the territory by yelling at the world's cartographers. It is possible that Australians will become less obnoxious if we all agree to stop discussing the fact of their obnoxious behaviour. But I would not bet on it. If in doubt, truth should be a sufficient justification for any statement purporting to claim a fact about the world in general or a model of causality in the world.
There are limiting cases where some statements might be irresponsible, like spreading information on how to make nuclear weapons from household items. In my estimation, those are pretty rare, however (actually, your view on how many statements ought be ruled as impermissible based on responsibility criteria is another way of phrasing the second requirement - you probably need a low filter here). There are also basic questions of politeness when it comes to not making unhelpful statements about a single person, particularly when made to that person. All of that applies. But outside of such personal interactions, there ought to be a strong presumption that truth is a sufficient justification for any statement.
This stops every argument descending into accusations about motives. The earth rotates around the sun, regardless of whether Galileo is saying so because of a devotion to scientific truth as he perceives it, or because Galileo is a contrarian rabble-rouser who likes to intellectually stick a finger in people's eyes, or because Galileo is intellectually committed to bringing down the Catholic Church. Truth is truth.
The third is that you don't take disagreement personally. If you think X, and someone else thinks Y, and X and Y are merely statements about how the world is, then we should be able to discuss this without the fact of my disagreeing with you causing you to get angry. If disagreement alone is enough to get you pissed off, then any discussion is a joint balancing of the strength and veracity of an argument, with my estimate of your current mood and the likely impact of the next statement on said mood. Such discussions tend to get exhausting very quickly for me. If disagreement, even about cherished beliefs, is not a source of anger, then we can talk about things.
Of course, you never quite know at first whether these requirements are going to be met. You try to feel people out about them.
But my experience is that with people who fit in these categories, I don't actually need any particular filter on what I say, although sometimes my remarks sound outlandish given popular sentiments. Usually, such people have a sense of humor about jokes on whatever the subject is too. They are worthy conversation partners.
In any case, if I do speak to you frankly, it is a mark of esteem, that I think you fit into all of the categories above.