Trust

Hi everyone,

I’m cheating this week. Yesterday was @threadapalooza, in which people assemble 100 tweet threads in order to stretch their ability to talk about a subject. I picked “Trust” for my topic. You can see the original Twitter thread, but this week’s letter is a slightly better formatted version of it.

1. I like Nguyen's paper "Trust as an Unquestioning Attitude" a lot, and will use a bunch of ideas from it. I'll explain everything from it that I need for this thread, but I'd also recommend reading it.

2. The idea that Nguyen puts forth in this paper is that there is a specific form of trust (he is open to there being other forms, but this form seems central) which is to adopt an unquestioning attitude towards something or someone, in a particular trusted capacity.

3. Importantly for this form of trust, it is always relatively specific and within limits. You e.g. trust a rope not to break under your weight, but you don't trust it to hold up a car. You might trust a coworker to do their job but not with highly personal secrets.

4. In Trust Beyond Reason I talk about how the process of developing trust works, and the relationship between trust and importance, and how you want to avoid scenarios where something or someone is important to you but you can't trust them.

5. One modification to the "Trust beyond reason" model is that trust is specific to a capacity, and importance is too, but the scoping is different. e.g. Making friends with a coworker can introduce a trust/importance mismatch you don't have in their professional capacity.

6. The reason this scenario is hard is that you coworker-trust them but don't friend-trust them, but their coworker-importance doesn't go away in learning to friend-trust them. Fucking up the friendship is higher risk because of that extra importance.

7. Of course, in the above I'm treating coworker-trust and friend-trust as totally distinct things, but they're not. Coworker-trust isn't just a monolothic "I trust them to be a good coworker", it's a whole range of questions that you don't feel the need to ask.

8. As a result one form of trust in a person can be a useful foundation for other forms of trust.

Essentially, there are skills of trusting someone for each someone and each capacity they might be in, and you have to learn each one but you don't have to learn them from scratch.

9. Now, of course, key to developing the skills of trusting someone in a capacity is that they have to be trustworthy in that capacity. Possibly more on trustworthiness later, but roughly something is trustwothy (in a capacity) to the degree that it is a good idea to trust it.

10. So, why trust? One answer is practical rationality. We are finite beings and can only keep track of so many things. If we are e.g. climbing and we are constantly worrying about the rope breaking under our weight, that distracts us from more important worries.

11. In this regard, trust is a coping strategy for the fact that we can't question everything all the time or we'd never get anything done. We get the reliability of something close enough to 100% and then round it upwards to 100% in our mind so we can get on with other things.

12. Another reason is emotional. Some questions the emotional cost of considering them is quite high, which makes it worth erring more on the side of trust than is strictly rational. e.g. I tend to assume my friends are honest with me to a higher degree than is likely true.

13. None of this is to say that if you ever question something you don't trust it. If I spot a trusted rope fraying I will question it. If a friend tells me the moon is made of cheese, I will suspect their honesty. But the unquestioning attitude is there absent external prompts.

14. Another way Nguyen highlights that trust can be regarded as a coping strategy for finiteness is that trust expands our capabilities in the world. That which we trust effectively becomes part of our agency, expanding the set of actions available to us.

15. Heidegger has the distinction ready-to-hand vs present-at-hand. e.g. when we are using a hammer to hammer something, the hammer is part of the way we are in the world, not an object distinct from us (it is ready-to-hand). If it breaks it becomes an object, present-at-hand.

16. We can only treat something as ready-to-hand to the degree that we trust it. An untrusted object is not something we can fluidly be in the world with, because we have to call its behaviour into question and treat it as part of the world rather than part of us.

17. We can see this dynamic with people as well as objects. Consider the difference between a coworker you can and can't trust to do their job. If you can trust them you can fluidly delegate their job to them, secure in the knowledge that it will get done. They are ready-to-hand.

18. One of the key features of trust is a sense of betrayal when the trust is violated, and this disruption of the ready-to-handness seems to be a key part of it. It's not just that our expectations are violated, but our very way of being in the world is disrupted.

19. This also can come across in the way that we trust ourselves. For example if our health fails us, we can feel betrayed by our bodies. We trusted our health (we didn't call it into question), and have been betrayed by that. Disruption of extended agency feels similar.

20. This sense of betrayal is partly the sense of reconfiguration - things we took as certain and thus foundational for how we learned to act have been called into question, and now we have to relearn a significant chunk of our behaviour.

21. I've been conflating trust in objects with trust in people, and this is the correct thing to do for understanding the notion of trust, but the notion of betrayal starts to tease apart one of the major ways in which they are different, which are our ethical duties.

22. For example, I trust my computer more or less. It's not that I think the computer is trustworthy in some absolute sense, it's that the cost of not trusting the computer is much too high. There may be specific ways I don't trust it that I account for but a lot I just trust.

23. When e.g. my display breaks, or I encounter some other technical fault, I feel this trust is betrayed.

What happens then? Well, I shout at the computer. A lot. I am not polite to it.

If a coworker betrayed my trust in the same way, I would not do that. That would be bad.

24. This is a relatively mundane difference - how we act towards objects is different than how we act towards people, even when the underlying emotion is the same - but the differences tease apart further. I might be *mad* at my coworker, legitimately, for the betrayal, but...

25. One of the big differences is there's a question of whose fault it is. Suppose your coworker has a specific habit of, say, taking lunch at the same time every day, and you use that habit as a prompt (This is Nguyen's example). They can change that habit without being culpable

26. The difference in trusting people vs trusting objects is that with trusting people there is a certain degree of negotiated social roles. Someone being trustworthy is not just a question of whether it's a good idea to trust them, but also of whether they endorse that trust.

27. Suppose I help you out, and you now trust me to always help you out in future. You ask me to help again and I say no I'm busy. You feel betrayed. Your sense of betrayal is not my fault, and your making it my problem would be bad behaviour.

28. This is not to say that *your trusting me* is bad behaviour - it might or might not be depending on what you do with that trust - but my role in that trust is different because I am a person with my own rights and agency, and you have to take that into account in trusting me.

29. Note that sometimes you trust people in the same way that you would trust objects because you have essentially turned their role into a black box - the fact that they are a person is not relevant to your trust, because you trust their role.

30. An example of this is that you would e.g. trust a cook not to spit in your food. This is more like you would trust an object - your trust would be identical if they were replaced with a cookbot - because their is no relational component to it.

31. This doesn't absolve you of ethical duties to them of course - if you treat a waiter like shit, don't be surprised when your food gets spat in - but the point is that the nature of the trust is slightly different. It's trust in them as a role rather than as a person.

32. Another different type of trust you can experience with people is trust that requires reciprocation because of the degree to which there is *shared* agency, where neither of you can be present-to-hand for the other without it being reciprocated.

33. e.g. consider dancing with someone. Dancing requires a level of trust in your partner's skill, and it is a level of trust that you cannot extend to someone who is constantly second guessing you - trying to anticipate your moves, correct your behaviour, etc.

34. This comes up in conversation too. If you are talking with someone who does not trust you, the conversation cannot be fluid because you are forced to constantly second guess what you're saying because you must very actively take into account how they will react.

35. This particularly comes up in deeply relational contexts - a thing that we experience with friends and romantic partners more than coworkers, who have at least some of the role element to their trust. There trust is integral to the entire relationship.

36. Deepening a relationship with someone requires deepening the trust between the two of you, and this is a process that requires a certain level of trust to be returned. You can only trust someone so far if you are constantly worrying they will decide you are untrustworthy.

37. This ties back to the "Trust as an unquestioning attitude" framing. One of the most important things in relationship development to be able to adopt this unquestioning attitude towards questions of the form "Do they trust me in this?".

(This is very hard)

38. This ties back into the notion of common knowledge of feelings I discussed in Common Knowledge of Feelings

A lot of this sort of trust is a common knowledge problem. I trust you, and you know that, I know that you know that, etc.

39. One complexity of the "unquestioning attitude" view of trust is that attitude is not really a unitary thing - we are capable of holding many thoughts, feelings, and attitudes simultaneously, and some of them will be more able to be unquestioning than others.

40. A big difficulty that comes up here is that trust requires a certain level of felt safety in the trusted thing. Suppose you have had a rope break on you - in future it will be harder for you to trust ropes regardless of how well you know the formal statistics on rope safety.

41. This is particularly problematic in interpersonal contexts, where we're just *full* of wrong beliefs about how safe people are that we learned in childhood. (cf. Your emotions are valid but probably wrong)

Emotions around trust and safety are learned in an environment that doesn't track adulthood.

42. This ties back to the problem of (4) - it is hard to learn to trust someone who is already high importance, because the risk factor of getting it wrong is so high. Bad news, that's the entirety of your childhood. Parents, teachers, etc. are very important.

43. The result is that adult emotions and negotiations around trust require a completely different dynamic than most people are emotionally prepared or have the skills for, and people treat learning to trust someone as much scarier than it needs to be.

44. I talked about the role of trust in creating safety, but equally there's a role of safety in creating trust. One way to not need to ask questions is that rounding to 100% I talked about, but the point at which you can do that rounding depends a lot on how bad things can get.

45. Suppose I'm like 90% sure things will be fine. Is that good enough to trust or not?

Well, it depends what happens if things are not fine. If I'd be mildly inconvenienced, that's probably good enough. If I'd die horribly, I'm definitely asking some more questions.

46. This is why to learn to trust before acquiring importance - high importance environments are unsafe because of the high stakes. If things go wrong, important things are lost. It's also why emotions around trust learned in high stakes childhood environments mislead adults.

47. It's hard to change our emotional lessons around this (ask me how I know etc) in a general sense, but one of the things we can do is to acknowledge the discrepancy between how much we feel we can trust someone and how trustworthy we actually think they are.

48. This doesn't even necessarily require acting differently if that's too scary (although I'll come back to this later for some cases where it might), just acknowledging the discrepancy and gradually using that acknowledgement to update the feeling.

49. When in this situation you can also use the role of safety in creating trust to try to build the feeling of trust appropriately by explicitly asking for affordances that make the relationship safer, at first small to see how that goes and then larger ones.

50. I think this trust/safety connection is part of why rich societies are often higher trust - a society in which everyone is safe is one where it's much easier to build trust because extending trust is lower risk for everyone and that tends to build on itself.

51. It's also why therapy works to some degree. You can trust your therapist with things that you would struggle to trust a partner with because if your therapist judges you harshly for them... oh well, they don't actually matter that much to you.

52. Another way in which trusting people is different from trusting objects is that although you never have a moral obligation to trust someone (you mostly can't have moral obligations to feel a particular way) you may have a moral obligation to *treat them as trustworthy*.

53. For example, suppose you grew up in a highly racist family, and you have since learned better but have not entirely managed to relearn the feeling. You may have more trouble trusting people of colour. There you do have a moral obligation to correct for that bias.

54. The problem is of course worse if you are actually just straight up racist, there by explicitly treating people of colour as less trustworthy you are actually behaving straight up immorally.

55. This is related to what the philosopher Miranda Fricker calls testimonial injustice (en.wikipedia.org/wiki/Epistemic…) - a lack of belief in someone's testimony for reasons solely rooted in their identity - but is a broader structural problem with how trust is assigned in society.

56. The distinction between trusting someone and treating them as trustworthy is I think quite important - you can have private questions all you want, but how you reflect those in your actions is something you are culpable for.

57. e.g. suppose I don't like a coworker and regard them more suspiciously as a result. I might harbour all sorts of secret doubts about them. Suppose I wonder if they're embezzling funds. So far, not culpable. Suppose I ask their manager that - suddenly a big deal.

58. This is particularly hard to navigate in close relationships, because they know you very well and so can probably tell how you feel, so the line between trusting them and treating them as trustworthy is a lot blurrier.

59. If someone you love treats you as untrustworthy, that hurts, especially if it's in ways you feel as attack on your ethics / character (e.g. if they react as if you would set out to hurt them). Unfortunately this tends to result in further relationship breakdowns.

60. As a result, a lot of relationship and communication work seems to be about untangling failures of trust and building structures that allow trust to grow. My impression is that this is made harder by the fact that most people don't even recognise that this is what's going on.

61. I've so far been treating as trustworthy as something you either are or aren't, but this is of course not the case. Most of the time trustworthiness is something you grow into. e.g. an apprentice is not trustworthy at their job, but becomes so partly by being trusted in it.

62. This is very true in a relational context, and ties in to the way that trust requires reciprocation. Suppose I have some problem that I want to talk about that sounds really bad on the surface. Are you trustworthy for me to talk about it? Depends on whether you trust me!

63. This example is particularly hard to navigate due to the specificity of trust - it may be that I am wrong about how you trust me, because you trust me in one context but not it turns out in the case in question. It's easy to misjudge how much someone trusts you as a result.

64. The "jumping to conclusions" thing is an interesting example of trust, because it slightly deviates from the "unquestioning" thing and is actually a case where trust practically *requires* you to ask questions, rather than assuming a conclusion.

65. I think this still kinda counts because e.g. the question that your conclusion answering is something like "Are they a bad person?" and the fact that you're assuming the answer "Yes" is because the trust gave you such strong prior attachment to the answer "no".

66. So a slightly more nuanced version of the unquestioning attitude view of trust is that it's more like the strength of your attitudinal attachment to a particular answer to the question - trust plus information that contradicts that causes you to look to clarify the situation.

67. Another thing that I think of as a moral obligation around trust - which is, annoyingly, in tension with the trust/treat-as-trustworthy one, is that you should reasonably reliably signal how much you actually trust someone.

68. I've had a bad experience in the past with someone treating me as if they trusted me and then very demonstrably not in situations where I feel like they very clearly owed me the benefit of the doubt. I'm still mad about this years later.

69. One of the reasons that I am mad about this is that they were treating me as if I were demanding that they trust me, when in fact what I was demanding was that they not lie to me about trusting me because if I'd known that they didn't I wouldn't have associated with them.

70. I think this is an important distinction to make. Hanging out with people who persistently refuse to trust you to the degree required by that relationship is exhausting and painful, and it's perfectly legitimate to choose not to do that.

71. It's like relationships. No individual owes you trust / a date, but it's still fine to seek out environments in which you are trusted / to try to find people to date, and it still sucks (and may be a sign of some sort of structural injustice) if you can't do that.

72. Switching topics a bit, one attitude where I think the unquestioning attitude model is quite helpful is for thinking about trusting yourself. I often end up recommending tactics to people that centre on removing the question of whether you can do things.

73. People often treat themselves as unreliable components in their own lives, especially if they have mental or physical health problems. This is literally a lack of trust in yourself, and treating yourself as untrustworthy is a fast track to feeling shame.

74. This creates a particularly unfortunate spiral where you treat yourself as untrustworthy, which creates shame, which creates aversion, which in turn makes you even more unreliable at the thing and thus even less trustworthy around it, and etc. This doesn't generally go well.

75. One way for this is to just admit you need help with it, delegate, and hopefully come to terms with that. Depending on the underlying causes of the problem this may be literally necessary. It's often helpful as a first pass even if it's not necessary.

76. The other path is to gradually learn to trust yourself, and simultaneously learn to be trustworthy for yourself. Generally I recommend doing this by gradually expanding the set of things you can do this with, starting small (to learn to trust your ability to build self-trust)

77. One thing you definitely *shouldn't* fully trust BTW is everything I'm saying in this thread. This is still an active area of work for me - both personally and intellectually - so I'm sure some of this is not quite right. Also I've run out of the well-understood bits now.

78. One useful tool for working on trust is to create environments where more trust is possible than would be outside these environments. e.g. I've talked about the support group I'm part of before. (cf. Notes on running a mutual support group)

The group agreement ensures certain trusts are possible.

79. One of the ways that trust manifests in conversation is a kind of... trust in the conversational structure? There is a question of "Is it OK for me to say this in this context?", and an explicit group agreement can help take anchor the answer to that question.

80. These sorts of explicit structures cannot make anyone who is fundamentally untrustworthy suddenly into someone you can trust of course - they're not for that. They're more for... allowing people to operate at the high end of their potential for trustworthiness?

81. In general the process of building trust and trustworthiness can be thought of as a process of aspiration, in Agnes Callard's sense (from her book Aspiration)

Aspiration is the process of rationally acquiring a set of values that you currently lack or imperfectly hold.

82. If you believe someone is trustworthy, but find it difficult to trust them, you *aspire* to trust them. Similarly if you are not trustworthy in a particular way but would like to be, you aspire to be able to reliably behave in the way that would allow you to be trusted.

83. One of the things Callard's book emphasises is the use of what she calls proleptic (anticipatory) values - things that you value as a kind of deliberate placeholder for the thing that you want to be able to value.

84. One example she uses is that of a mentor, whose opinion you value as a stand in for learning to value the thing the mentor does.

I think you can use trustworthiness as this kind of proleptic value for trust, by explicitly valuing it as a guide to learning to trust.

85. How does this work? Uh, I'm partly still figuring that out, sorry, but one thing that I think helps is to continually operate slightly outside of your comfort zone and make it explicit that that is what you are doing. Creating safe external structures to enable that can help.

86. One thing that I think works well as this kind of proleptic value for trust is the success of relationships. Pretty much every relationship has difficulties that are bottlenecked on trust, because there's no such thing as unconditional trust and relationships are a big deal.

87. Of course it "works well" in the sense that it's a terrifying ordeal of high risk / high reward. Fortunately it also works well as a proleptic value for trustworthiness, which is a lot less scary to work on. cf.Being safe for others

88. This doesn't just apply to romantic relationships, or even friendships, it also applies in organisations. Everything works better when you can trust your coworkers, and using organisational success as a stand in can in theory help you build trust and trustworthiness.

89. Bad news though, it's basically impossible. The power differential is too large. It's hard to impossible to be trustworthy to someone you have that much power over, because the amount of damage you can do is so large that it requires an incredibly high bar to fully trust you.

90. This suggests that things that tie a managers' hands over how much they can screw over employees would actually produce more functional companies. I suspect it doesn't cleanly work out that way because of the emotional legacy problem, but maybe?

91. An interesting dynamic of the "Trust as an unquestioning attitude" thing is that there is a sort of... metatrust? A kind of trust in the potential to trust someone, whether you approach them with an attitude that trusting them might be possible or not.

92. One of the factors that goes into metatrust is a kind of level of social connection. A drive-by random on stranger it's probably not worth learning to trust them, so we rarely extend benefit of the doubt. A coworker it's worth finding out if they're trustworthy.

93. There is also a personal aspect. If you have deep seated trust issues (don't we all) then your level of metatrust is going to be low - you'll at best question whether it's worth learning to trust someone, at worst assume that it's not.

94. This suggests that even though trust is specific to a person and a context, learning to trust someone in one context increases our trust in it being worth learning to trust them in other contexts, and also in being worth learning to trust other people in that context.

95. It also suggests that learning to value trust itself is a useful way to learn to trust more, because if trust is valuable it's more worth finding out if it's possible. Thus *valuing trust itself* is a proleptic value for valuing trust in any particular context.

96. It also suggests that understanding the dynamics of trust is itself a useful way to build trust, because "worth it" is also dependent on your ability. If you were literally unable to trust someone, it would not be worth trying to find out if they were trustworthy.

97. So, for example, reading a 100 tweet thread about trust and why it's good and how it works might be a useful way to build metatrust, and thus to get better at trusting people. Just saying.

98. I'll close by noting that trust isn't a special interest for me because I'm naturally good at it. I know full well that in asking you to trust more people and trust them more I'm asking something incredibly hard. It took me until my 30s to really be able to do it.

99. I'd say it's really only been the last few years where any approach other than lowering the risk to the point where it didn't matter if I was wrong worked. I got very good at that, that's why I'm good at saying things in public that other people would find too vulnerable.

100. But everything works better when you surround yourself with trustworthy people and learn to trust them, and when (and this is even harder) you learn to trust yourself. The world is better, and larger, and you feel more secure in it. It's worth it, I promise.

This concludes my thread. I trust that you have learned something from it, if only that I really do overthink things enough to try to explain romantic relationships in terms of Heidegger.