Post 35: Your Standards are Too High — Neel Nanda

Introduction

I’m a perfectionist, and a pretty neurotic person, so a common experience for me is feeling dissatisfied and guilty. Some part of me is deeply convinced that everything should be easy, and fast. That if it’s not, I am failing. That I could have done better.

When I do an exam, no matter how well I actually did, this part is always drawn to the marks missed and the dumb screw ups. When I'm thinking about something hard and find an inspired idea, it always feels obvious in hindsight and I kick myself for missing it. When I'm tired and spend the first hour of a day procrastinating, even if I completely pull it around from there, I'll compare what I do to what could have been.

I think this is an extremely common pattern, and one that I see in many of my friends. And the crux of it is the word could have, the standards we hold ourselves to. This mix of guilt, insecurity and unrealistic standards is a major loss of productivity, and a major source of unhappiness. And this pisses me off.

I am going to spend this post trying to argue exactly why this is so terrible, and how I personally think about doing something about it. I expect that some lucky people reading this won't relate at all to these problems, and more power to you! But for those who resonate, I hope my thoughts can add a useful perspective!

The key framework I will argue for is as follows:

  • My coulds push me towards things that are fundamentally unrealistic, and my intuitions are badly calibrated. Hindsight bias, planning fallacy, far mode thinking all combine to create a fundamentally wrong set of expectations
  • The point of standards is to push me to be better. They are an instrumental tool, not a terminal goal. There are ways my standards do not help me to become better, and so my desire to be better means I need to change them
    • The impulse to become better is not the problem, it's the execution
  • It's hard to change your standards just by willing it, or by realising they're unhealthy. My main tool here is to become better calibrated, and try to ground my standards in the true goal of becoming better.

There is also a problem of having bad goals, and standards tying me to goals I no longer ultimately care about. This is a very real problem, but I won't focus on it here - my earlier post on prioritisation somewhat gives my thoughts there.

It's easy for these discussions to get somewhat abstract - to agree that realistic standards are good in principle, but that it would never work in practice. To counteract this, I recommend taking a moment right now, to identify a recent time where you didn't meet your standards and felt guilty about it. And to apply all of my ideas to that specific example, and see if they resonate.

I will personally mostly focus on the case of productivity and focus, since that is important to me and highly salient, but I think these ideas apply far beyond that - judging my talents and comparing myself to others, life satisfaction and a sense of progress, and more generally anything where insecurities dominate. And I hope these ideas are of value there too.

My standards are unrealistic

The crux here, is that the only thing I can change are my future actions - actions taken under uncertainty, where I must apply judgement and balance trade-offs. But my standards judge me by past data, with all of the benefits of hindsight.

And it is exceptionally hard to look past hindsight and judge the past from the perspective of how it seemed at the time the concept of hindsight bias. This is a fundamentally difficult problem, and doing this well ever is a struggle. It is utterly unreasonable to expect our intuitions to be capable of doing this well on their own!

This is especially bad when looking back on any kind of conceptual progress, clever ideas where all of the work came from identifying the idea in the space of all possible ideas. I see this a lot in maths - group theory is a beautiful area of maths, that took decades of effort and iteration to distill into its modern form. Yet, when I think about the idea of how I'd characterise a symmetry, the framework of group theory feels obvious! And this problem recurs again and again when looking back on any intellectual work I've done - mistakes I've made, rabbit holes I've gone down, clever solutions I've found after a while of staring at a problem. This is a fundamentally hard problem, and to get anywhere I need to record how things felt in the past.

Further, this is actively damaging! There are mistakes I've made in the past, things I could have done better, ways I can learn and grow stronger. The guilt pushes me to flinch away from these, rather than examining my past shortcomings in detail, seeing what I can learn from them, and feeling motivated to put in effort to do better next time.

A further problem, is what I call the floating ball of rationality problem. I conceive of my actions and feelings as being under total conscious control. If I don't keep working until I drop, this is a failing. If somebody annoys me and I fail to utterly brush it off, this is weakness. And clearly it is psychologically impossible to just will myself into these states of mind - if they're even possible, it takes meaningful work and effort to get there. But my sense of what is easy is broken.

Another problem is the planning fallacy, and more generally far mode thinking. When I think on future projects and plans, I see them from a zoomed out view. I don't account for dumb mistakes, distractions, tiredness and procrastination, even though this always happens. I don't account for things being harder than expected. And this means I am consistently disappointed with what happens.

Worse, even when I do do something well, I rarely feel satisfied. I gave a talk last week, and I'm super happy with how it went and what I learned in making it. And, including background reading etc, it took about 20-40 hours to prepare and give. Extrapolating out, this would give me about a talk a week, which I feel super satisfied about. Yet, if I actually think about what I did in that time (and even now while writing this post about how this is dumb), it feels like it should have taken 10 hours tops. Guilt and perfectionism is trying to spur me to do awesome things, but sometimes I do things that I'm proud of, and I deserve to feel satisfied when that happens

Complacency and compassion

At this point in this conversation, some part of me starts to feel concerned about complacency. I know that my standards are unrealistic. But they also serve as a big part of my motivation system. And I feel concerned that if I just removed them, that I'd be complacent, and never do anything. And that feels terrifying. I know that high standards make me feel less happy, and that it'd be awesome if I could be productive and feel awesome about everything, but that doesn't feel possible.

A common counter-point is the idea of self-compassion. I find it most compelling to think about this in terms of how I'd advise a friend in the same situation. I feel very confident I'd think the friend was being way too hard on themselves. I'd encourage them to zoom out, relax a bit, and try again a bit later. That guilt, pushing and forcing yourself is rarely effective or sustainable. This discrepancy is clearly an inconsistency. And, on reflection, I feel pretty satisfied with the advice I'd give to the friend.

Here, I find it helpful to disentangle the guilt & standards from the desire to do something awesome and well. The standards are purely instrumental, and I have no attachment to it for its own sake. And it generally feels pretty clear that if I can get intrinsically motivated, this is a massive upgrade! The times when I have felt excited about what I'm working on and in flow have been some of the most productive times of my life. So the question is whether it's possible to move from this. (And if the idea of intrinsic motivation resonates with you, this post gives my thoughts on how to cultivate it).

As for the fear that relaxed standards will lead to complacency - I strongly relate to the deep, neurotic desire to always be in control of things. And a fear that relaxing will lead to a loss of control. But I don't think this is actually true. A thought experiment: You have a big red button in front of you. If you press it, it'll remove all of these feelings of guilt. You won't work on the task any more, but will feel totally fine about this. Do you want to press it? I certainly don't - there are much deeper reasons I care about what I'm doing! And those deeper reasons will still be there, and can still keep me focused, without the blunt hammer of guilt. Guilt is a tool, not the only tool. (And if you would press the button, are your perceived goals your true goals?)

Another framing: Guilt is a form of negative reinforcement, and satisfaction is positive reinforcement. Over time, we learn to avoid negative reinforcement, and to seek positive reinforcement - at our heart, we're all reinforcement learning agents seeking reward. And getting a constant negative reward prompting us to keep striving and becoming better can work, if enough work would actually help us become awesome and worthy. But it's much easier with a shaped reward - where sometimes we feel satisfied and happy with things. This creates much tighter feedback loops, and makes it much easier to point yourself in the right direction, measure progress, and keep motivated.

Another thought experiment: Suppose you woke up tomorrow, and you were consistently twice as productive as you have been up to today. Try to really imagine it. How surprised do you feel at this picture?

I'd feel pretty surprised at this - progress tends to be continuous, not discrete! It's much more likely that I get slowly better, day by day. And so, to motivate this, it suffices to motivate marginally better performance. If I set my standard of satisfaction at, say, the 60th percentile of days, this is great! It's a shaped reward, pushes me in the right direction. But also, the only way it could feel too low is if I consistently hit it. And that's an awesome outcome. The progress of becoming great is for every day to be marginally better than the last. This is exactly what the kind of realistic and helpful standards that genuinely push me to better would look like - accurately tracking my progress and how good I currently am, and pushing myself to be marginally better than that.

This has two important takeaways:

  • Sometimes things don't go great! You fall below your standards. This is sad - often you can learn something from it, and you want to minimise it.
    • But it's also completely to be expected! See what you can learn from it, and move on. The process of becoming better is full of random noise, and will have off days
  • Your standards need to be calibrated. You're still pushing yourself to be better, you still have standards. There is a notion of success, and a notion of failure, and this should be enough to stave off the fear of complacency. But your standards should be marginally better than your current performance.
    • If my current standards were realistic, my rate of growth would be far beyond the bounds of plausibility, so clearly this is not helping!

Think back on the last 2 weeks, and how often you felt satisfied with what you got done that day. If you're anything like me, the number is way less than half. Is this the kind of reward that's going to best help you learn and become stronger?

(I'd highly recommend Nate Soares' Not Yet Gods, for another perspective on the same theme)

Becoming calibrated

So, the underlying goal is to become better, to be motivated, and to slowly, incrementally improve. And this needs standards that are accurate. That are well-calibrated to my true abilities. I have a few thoughts about how to achieve this:

  • Collect data!
    • Far and away the best way to overcome the planning fallacy is by taking the outside view - how long does this kind of task normally take?
      • Roughly track your time, and how long different projects take. And then, when starting a new project, use this number as your baseline next time
    • Software like Toggl can be valuable for systematising this time tracking, though personally I find it most useful to think in units of 50 minute pomodoros
    • Blogging has made the value of this approach feel extremely visceral to me. I know that I write at about 20-30 WPM (including planning), and this is my anchor. And the posts always feel like it's going slowly, like I could be doing way better. But at the end, when I look back on how it actually went, it's normally pretty typical - the emotion in the moment is just giving me garbage data
    • I highly recommend having some kind of data collection system, eg a Trello board to track time spent per project/project chunk, or a spreadsheet. You need to actually have numbers to look at, don't just be lazy and think you're keeping track mentally
  • Notice awesome days. The days where you were focused throughout, 90th percentile days. Where everything went right, you feel satisfied with what you actually got done.
    • Notice and introspect on the feeling of satisfaction, and then record what you actually got done
    • And the next time you feel dissatisfied with your speed and rate of progress, extrapolate out what you think you "could have" got done, and compare it to this awesome day.
      • Better, try to picture a day where you actually did what you think you might have. Disengage far mode, and try to picture the details. Introspect on this. And ask, would I be surprised if this day had happened?
    • If you're anything like me, your standards expect days way better than what actually happens, and this can help get surface area on it
  • Other-ise while beating yourself up, imagine somebody else in your situation, and imagine what advice you'd give to them
    • I find this extremely useful for overcoming insecurity and for seeking upside risk. Eg, if I'm considering applying for a job, I have super high standards. I need to feel confident I'll get it, that I'm not overreaching or pretentious, that I've done my homework. Yet, if I was advising a friend, I'd point out that the downside is tiny and they should just go for it!
    • I have a similar tendency to be strongly averse to bothering other people or incurring costs on them, while having no such compunctions to costs on myself, which obviously leads to suboptimal outcomes (and, eg, means I utterly suck at asking for favours). Other-ising is great for addressing this too, and tends to engage much more competent parts of my brain
    • Related technique: Imagine a specific friend in an analogous situation. Think about the advice I'd give to them. And then try to explain why that doesn't obviously also apply to me
      • I find this can further help to break through BS forms of insecurity
  • Ask other people! Ultimately, when I get too stuck in my head, the best way to get an outside view is literally to ask a friend for advice!
    • Remember - while your friends can be wrong, it's much more likely that you are wrong. Your goal is to find truth, and their beliefs are likely more entangled with the truth than your's. You have a clear bias, and need to counteract it.
    • Another concern: that your friends are actively lying to spare your feelings and can't be trusted. I try fairly hard to be good at taking criticism well, and to set clear norms of directness and honesty with my friends, so I trust them to be honest in cases like this. I expect this would be less effective without those norms. (Though those norms are great, and I'd highly recommend!)

Further, I think that poorly calibrated standards are much easier to fix when you directly notice them, but harder to fix when you're tired or paying less attention. And so I've found this extremely valuable to systematise, to aim for useful standards to be a default. My current system:

  • I have a Trello board where I track my current projects (a card should represent 1-10 hours - if more than that, try to break it down)
  • In a weekly review, track how long I actually spent on each project this week, figure out what I want to work on this week, and estimate how long each will take. Always use past data to make these estimates, if possible
    • If I was badly off on a prediction, try to figure out why this happened, and what to learn from this!
      • Eg, the lesson that novel tasks always take longer than anticipated!
    • A useful psychological hack: Estimate the time a project would take. Try to be pessimistic, because planning fallacy. Then, imagine you needed to wager $1000 that you'd have the project done in that time. Record both times, and then calibrate at the end. When I do this, the $1000 time is often a better guess...
  • Then, at the start of each day, look at my project board, and figure out what I want to do today. Look at how many pomodoros I have, and try to allocate tasks across pomodoros. Write down this priority ordered list in Complice, and make the default action to follow it (with some leeway when my planning inevitably sucks)
    • This works best as part of a morning routine, or the night before, rather than at the start of work. This creates separation of planning and doing - it's easier to work on things with a sense of high level direction, and it's less aversive to plan without the imminent prospect of needing to do the task
  • In each pomodoro, use a Google form to write down concretely what I plan on doing, and to attempt to quantify progress. And at the end, record what I actually did.

This system is somewhat overengineered, and has moderate overheads. But I've found it valuable - the tiered system creates a lot of tight feedback loops for predictions, and I've become much better calibrated about how long things actually take me. If your calibration also sucks, I'd highly recommend trying a similar strategy yourself, at least for a few weeks!

Exercise: What would you do as part of an ideal calibration and standard setting system? Where could these fit into a routine, such that it feels like the default action to follow? What feels most missing from how you currently do things? And what could you do right now, to implement such a system?

(I'd highly recommend Lynette Bye's Measuring Progress, for another perspective on the same theme)

Notice your anchor point

Even if I can perfectly calibrate my standards, this doesn't solve everything. Sometimes I'll have an off-day, where I just fall short of a reasonable standard. Sometimes I'll spend my first pomodoro procrastinating and get nothing done. And often, this throws off my entire rest of the day. I feel obliged to make up for the lost time, to recover the time lost. And this is obviously ridiculous! If my standards are well-calibrated, it's unreasonable to expect my future self to now exceed them.

The fundamental problem here is that the goal of standards is to shape my future actions for the better. Yet, my standards are anchored to my past actions.

And this can often be terrible! I feel off, I know there's no way I'll feel happy with the day, and so I just give up. I wanted to be in bed by 10pm, it's now midnight, so I've already stayed up too late - what's a few hours more? This is the phenomena Nate Soares called failing with abandon, and it is both insidious and extremely unhelpful.

The underlying problem, is that guilt and obligation are blunt, unnuanced tools. They can be useful, but don't automatically adapt to new situations - I call this phenomena being anchored. This is very much worth watching out for and trying to resolve - if you're feeling miserable in a way that isn't helping you to become stronger, that's just obviously terrible.

My main tool for resolving this specific case, is to un-anchor zooming out, reflecting, becoming centred and re-orienting to a new plan for the new situation. This can often feel aversive, because I'm already behind and can't afford to waste any more time! But this is clearly dumb. If I'm being unproductive and meandering, this is now the default path. Pressing the Try Harder button isn't going to pull me from that path. While doing something different, zooming out and adapting can change that path. True competence isn't avoiding problems, it's being able to recover when they inevitably arise.

Finding the right goals

A similar phenomena occurs when your goals are wrong. You feel obliged to go to lectures and learn, because that's what a student does, even though you're learning poorly and don't care about the material. You feel obliged to stick with your current project and work through it, even though it's ceased to be fun or meaningful to you - there's the drive to not leave things unfinished. The drive to finish the book that became dull at chapter 5. This is fundamentally a problem of your standards being anchored at the wrong points.

It is valuable to notice this and be aware of this, though difficult to fix, and I won't dig into it too deeply in this post. I've gotten some mileage out of internalising that the world is fundamentally full of trade-offs, and that I need to learn how to prioritise. And by cultivating a sense of outrage at standard programming cached in my head. That I've absorbed these thoughts and obligations, they are not under my conscious control, and that this leads to bullshit aversions holding me back from my true goals. And this helps motivate me to do something about it.

Sometimes the solution is to calibrate your standards better, as I've outlined here. But sometimes the solution is to scrap your standards entirely, and re-orient to your true goals! And it's worthwhile not to confuse the two.

Conclusion

In conclusion, I think a ton of unhappiness, akrasia and lost productivity comes from poorly applied guilt and poorly calibrated standards. This is extremely common, and terrible.

We have a lot of biases that mean our standards are badly calibrated, unrealistic and dominated by hindsight bias. It is valuable to have realistic standards - intrinsic motivation is far more effective than guilt, and tight feedback loops are far more motivating and better at pointing us in the right direction. This is obvious if we imagine advising another on overcoming these problems.

The goal isn't to have unrealistic standards, it's to have standards slightly above what I typically achieve - the drive to improve and be better is a valuable one and to be cherished, but my default approach is ineffective and can be improved. Ultimately, the goal of standards is to improve our future actions - the past is just training data. And this is the true goal to be grounded in.

And getting well-calibrated standards is a tractable problem. Progress can be made here by making predictions about how long things will take, recording data, taking the outside view and building this in as a systematic part of my workflow. And it is possible to become better calibrated over time. You have a bias, and being aware of a bias doesn’t dissolve it. But it can be overcome.

I've mostly focused on productivity and workflow here, as that's the aspect I've thought about the most, but I think badly calibrated standards are everywhere in my life. And these techniques of collecting data, seeking advice, being grounded are powerful parts of my general toolkit for dealing with guilt and insecurity. I find it hard to suppress instincts like that, since the desire to do well and seek excellence is precious to me. So I strongly resonate with a solution centred on digging deeper into that desire.

And if the ideas in this post have resonated with you, I'd urge you to do something about it? When was the last time you felt really satisfied with what you've gotten done? How often do you feel guilty? And, if you were designing your mind to help you become awesome, what would those answers look like? Are you happy with your mind working like this? And if not, what can you do about it?

(If this framing of guilt and obligation as things holding me back resonated, I'd very highly recommend the Replacing Guilt series by Nate Soares - they're some of the most valuable things I've ever read about guilt and mental health)