Jiminy Cricket Must Die

Before I get on to the main point of this post, let me ask you a question: when reading a piece someone wrote, does it matter if there use of language conforms to the rules of grammar your used to, or is it acceptable if their writing an they’re own dialect?

If you’re like me that sentence made you uncomfortable1. There’s a kind of twitch, a feeling that something isn’t right, a desire to fix it. Right? It feels wrong.

If you’re a Python programmer, I’d encourage you to go look at should-DSL and stare at the examples for a while until you understand how they work to get a similar sense of wrongness.

In his book “Voices: The Educational Formation of Conscience” Thomas Green describes these as voices of conscience. He defines “conscience” as “reflexive judgment about things that matter”, and argues that these voices of conscience are not intrinsic, but learned as part of our moral development through our membership in communities of practice – English speakers, Python programmers, etc.

That is, nobody is born knowing how to write Python or speak English, but in the course of learning how to do so we also learn how to behave as English speakers and Python programmers. By our membership of these communities we learn their norms, and by those norms we acquire voices of conscience that tell us to follow them. Because we exist in many different contexts, we acquire many different voices of conscience. Often these may disagree with eachother.

Importantly, we can acquire these voices for the norms of a community even if we don’t adhere to those norms.

Green distinguishes between three different ways of experiencing a norm. You can comply with it, you can be obedient to it, and you can be observant of it. Compliance is when your behaviour matches the norm (which may e.g. be just because it is convenient to do so), obedience when you actively seek to follow the prescriptions of the norm, and observance is when you have internalised the norm and experience it as a voice of conscience.

To this list I would also add enforcement – whether you try to correct other people when they fail to comply with the norm.

It’s easy to construct examples that satisfy any one of these but not the others, but for example the sentence at the beginning is an example of non-compliance where I am still observant of the norm: I know the sentence is wrong, but I did the wrong thing anyway. Similarly, I am observant of the norm when I notice that other’s usage is wrong, even if I make no attempt to enforce it (which generally I don’t unless I’ve been asked to).

It is specifically observance (and to some extent enforcement) that I want to talk about, and why I think the voices metaphor breaks down.

Let me turn to a different source on ethics, Jonathan Haidt. In his book the righteous mind he presents Moral Foundations Theory, which proposes a set of “foundations” of ethics. I think Moral Foundations Theory is approximately as useful a set of distinctions as Hogwarts Houses2, but a lot of the underlying research is interesting.

The following is a story that he presents to people as part of an experiment in where morality comes from:

Julie and Mark, who are sister and brother, are traveling together in France. They are both on summer vacation from college. One night they are staying alone in a cabin near the beach. They decide that it would be interesting and fun if they tried making love. At the very least it would be a new experience for each of them. Julie is already taking birth control pills, but Mark uses a condom too, just to be safe. They both enjoy it, but they decide not to do it again. They keep that night as a special secret between them, which makes them feel even closer to each other. So what do you think about this? Was it wrong for them to have sex?

Jonathan Haidt, The Righteous Mind, 2013, p45

If you read interview transcripts of people’s reaction to this story (which was deliberately constructed to provoke a disgust reaction), the common factors that emerge are that it feels wrong, and to the degree people can justify it they first of all struggle and then eventually do so on the basis of it being a norm violation rather than being able to point to any “objective” reason why it was wrong (partly because the story was specifically constructed to achieve that – the parties are consenting adults, there is no risk of pregnancy, no harm is done, the story is kept a secret so does not normalise something that might be harmful in general even if it’s not in the specific case, etc.). People make their judgements about morality based on intuitive, emotional, responses to the scenario and then try to reason their way to that conclusion.

It is useful here to have the distinction between a belief and an alief. A belief is something that you think to be true, while an alief is something that you feel to be true (they’re called a-liefs because they become b-liefs). Haidt’s research suggests that aliefs and not beliefs are the foundation of people’s moral “reasoning”.

This then is the source of my disagreement with Green’s metaphor of a voice of conscience: Conscience doesn’t start with a voice, it starts with a feeling. There is then a voice on top of that, allowing us to reason about and navigate the norm, but the feeling is what gives the voice its power. Without a felt sense of conscience, the voice is just knowledge that this is a norm that should be obeyed if there are not to be consequences. Once the consequences go away, so does obedience to the norm, but if you have learned the feeling of conscience then it will linger for a long time even if you leave the community where you learned it.

How do we acquire these norms (Green calls the process normation)? Bluntly, operant conditioning.

When we obey the norm, we are often rewarded. When we disobey it, we are often punished. Sometimes this is enforced by other people, sometimes this is enforced by reality, sometimes this is enforced by our own latent fears about standing out from the crowd (itself a feeling of conscience that we have acquired – standing out makes you a target).

The conditioning trains our habits, and our feelings, to comply with the norm because we learn at an intuitive level to behave in a way that results in good outcomes – behaviours that work well are repeated, behaviours that work badly are not, and we learn the intuitive sense of rightness that comes with “good” behaviour from it.

So what does conscience feel like? Conscience feels like following the path in the world that has been carved for you via your training. When you stick to the path, it feels right and good, and as you stray from it the world starts to feel unsettling, and even if you no longer fear being punished for it you have learned to punish yourself for it.

This may sound like a very cynical framing of it, so let me just reassure you that I am not about to start talking about “slave morality” and how we should throw off the oppressive shackles of society’s unreasonable demands that we should be nice to people.

But there is an important point in this cynicism: The process of conscience formation, and the feelings that result, are morally neutral.

The ways that we learn the rules of grammar are the same as the ways in which we learn that harming people, are the same ways that people learn that, say, homosexuality is wrong. We might learn these through our memberships of different communities, and we certainly learn them with different strengths, but we acquire them through broadly the same mechanisms and acquire broadly similar senses of rightness and wrongness through them.

Over time, we are part of and pass through many communities, and accrue senses of conscience from them. Because of the shared felt sense of conscience, we cannot necessarily distinguish between them, and we end up with an underlying muddy sense of rightness and wrongness with no clear boundaries or sense where particular parts come from. Some things feel like the sort of thing you’re supposed to do and some things don’t.

Much of this underlying sense of conscience is bad and needs to be destroyed.

Because the formation of conscience is a morally neutral process, the consciences that we form may be bad.

How often does this happen? Well, consider this:

  1. We learn our consciences from the groups in which we are embedded.
  2. We are all a part of society.
  3. Society is really fucked up.

As we go through life, we pass through different spaces, and learn their norms, and then when we leave we drag the consciences that we learned there along with us. Many of these spaces are broken, and we learn bad norms from them that we have to break out of if we want to grow. e.g. “men shouldn’t express their feelings” or “I’m not allowed to set boundaries”.

As we grow more morally sophisticated (which is not necessarily the same as more moral) we come to understand that there is a distinction between “feels wrong” and “is wrong”, and that just because we react to something with visceral disgust doesn’t mean we should necessarily consider it immoral.

As a result we separate ourselves from the feeling of conscience and privilege the voice of conscience. If we can’t explain why something is bad, we say things like “Well I wouldn’t do it but obviously that’s fine if you want to”. Sometimes through gritted teeth, sometimes while genuinely managing to convey the impression that you should do you.

At this point we pat ourselves on the collective backs for having become enlightened and moved past those ugly primitive urges and archaic social constructs. We’ve moved from listening to the feeling of conscience to the voice of consicence, and because voices are the tool of rational analysis we think we have moved to a higher level of moral understanding.

We haven’t. This was the wrong thing to do. It is at best a transitional step, but more commonly it is a dead end.

The problem with this strategy is that it conflates enforcement with observance, and aliefs with beliefs. We think that because we have stopped enforcing the norm we have stopped observing it3, and we think that because we no longer believe something is immoral we no longer alieve it.

It is important to bring our moral beliefs and aliefs into alignment, but the way to do that is not to suppress our feelings on the subject. Suppressing feelings doesn’t make them go away, it buries them and increases the associated trauma. Disassociating from ourself isn’t the path to becoming a more moral person, it just causes us to build up trauma in the areas that we’re ignoring.

If we want to continue our moral development past a certain point, we need to learn the skill of navigating our conscience, and that means getting to a point where we can put aside the voice of conscience and look deeper as to where the feeling comes from.

Instead of flat out declaring that our moral intuitions are wrong and we shouldn’t feel that way, if something feels wrong we need to be able to approach that feeling and ask questions of it. The most important questions being where, why, and from whom did I learn that this was wrong?

This doesn’t make the feelings go away. Making the feelings go away is not the goal, at least initially. The goal is to get a handle on them, and to get to the point where you can make progress. Rather than being in constant conflict between voice and feeling, you can start to navigate between the two, and hopefully eventually come to a healthier position where your moral aliefs and beliefs line up.

This is extremely hard. I wish I had nice, easy, shortcuts to offer on how to do that, but I don’t. I’m still figuring all of this out myself.

I think the basic starting point for solving this is getting better at engaging with our feelings at all. I’m still processing how best to do that, and I owe you one blog post about this, but for now I’ll hand you over to this Twitter thread I wrote yesterday.

Once we’ve done that, the next step is to do the work. Ask ourselves questions, learn why we feel a particular way. If we still think the feeling is valid afterwards, that’s great, we’ve learned more about our moral intuitions. If, on reflection, we decide that actually this moral intuition comes from a bad place, we can start to examine it and, from a place of safety, start to unpick some of the moral damage that was done to us.

This entry was posted in Uncategorized on by .

The missing social technology sector

There’s a thing that I’ve been puzzled about for a while. There’s an entire sector of industry that, as far as I can tell, does not exist and should. For want of a better term I’m going to call the thing that sector would do social technology1

It’s entirely possible that the answer is that this sector does exist but I’m not in the normal target market. If you’d like to come along and shout “You idiot tech bro. X, you’ve invented X.” then honestly this time I’d be delighted because I legitimately have tried to solve for X and failed.

What is social technology?

Social technology is “technology” built out of groups of people following rules, maybe with assistance from simple props, rather than software, machinery, etc2.

What does that look like? Well, let me give you a free sample. Here’s how to have a great high level conversation with someone (a friend, coworker, etc).

  1. Find yourself a five minute sand timer (you can use a phone timer if you like, but I find the soft limit of having it be visual is helpful rather than a loud beep).
  2. Place the timer in front of one of you. That person picks a topic they’d like to talk about. Maybe write down a sentence about what it is if you like.
  3. Now set the timer going and talk about that topic, trying to stay more or less on focus.
  4. When the timer runs out, the person who it is not in front of can either turn it over and leave it where it is to continue discussing the topic, or claim it. They now get to set the topic as above. Often what they will say is “I’d like to focus on X that we were talking about” or “What you said about Y reminded me about Z that I’d like to discuss” but that’s not required.

This is from a discussion system I’m vaguely tinkering with called TickTalk. I’ve only run this two player version once, but it seems to work very well, and results in great conversations of a particular sort. Certainly the multiplayer version has been a great success.

What makes something like TickTalk a technology?

It’s a technology because it has to be invented and developed like a technology, and it has results like a technology: Someone had to come up with the idea, the idea and practice had to be refined through use, and it provides us with capabilities greater than what we had prior to its invention.

Sure, TickTalk is just a system of rules, it doesn’t feel like a thing per se, but software is also just a system of rules, and we consider that technology to the point that “tech” is almost treated as synonymous with software these days (although it shouldn’t be).

What’s interesting about social technology?

Social technology has several key interesting features that I think distinguishes it from most other technologies.

Firstly, it is built on extremely powerful components. Humans are incredibly flexible compared to most things we build technology out of. They interpret, they adapt, they fill in the blank spots. If you study the literature around complex systems, essentially every complex system that works does so because of humans running around behind the scenes papering over the bits where it doesn’t work and filling in the details. If you imagine programming with a “do what I mean” construct, social technology development isn’t far off. Social technology lets us jump over all of the things we don’t know how to do with computers but do know how to do with people. We could have been building artificial superintelligences thousands of years ago, but have still mostly failed to do so.

Secondly, social technology as I have defined it is incredibly cheap. Although some of it (e.g. therapy) admits and requires a lot of deep expertise, many interesting social technologies can be written down on a page, understood with one run through, and implemented based on everyday household items.

Thirdly, social technology is incredibly adaptable. Because humans are flexible and social technology is cheap, it’s essentially free to take pieces of social technology apart and put them together again in new configurations.

Finally, I think social technology is interesting because we can use it to solve real and pervasive problems with our lives. The fact that it’s cheap and adaptable means that we can start to deploy it everywhere almost immediately by looking at areas of our lives that could use improving and finding a way to slot it in. TickTalk has been immeasurably helpful in improving my life over the last year because of some of the conversations it has afforded, and I think there are many more opportunities like that available to people and currently not being taken.

Why do we need a social technology sector?

Social technology exists today. It’s certainly not a thing I’ve invented. What I find puzzling is that it does not seem to exist as any sort of unified field, and does not seem to have a substantial amount of industry associated with developing it.

There are a number of areas which are essentially social technology subfields. For example:

  1. Liberating Structures are a type of social technology for management and business.
  2. Social Choice Theory is a weird thing in that it is an elaborate theoretical study of how social technology for decision making might work but largely skips out on actually doing the work to go from science to technology (e.g. there’s no user experience work). Voting Systems are very much a social technology.
  3. Pen and Paper Roleplaying Games are an entire thriving social technology games sector that is largely disconnected from any broader practice. Maybe this will change with the rise of the professional dungeon master, who knows?
  4. Many methodologies (agile, lean, etc.) are arguably a type of social technology3.
  5. Bureaucracies are enterprise social technology deployments.
  6. A lot of social sciences can be kinda regarded as studying the social technology we implicitly build.
  7. Therapy is a social technology (or, really, a large family of social technologies).

So there’s certainly no absence of social technology in the wild, but it mostly seems to be very ad hoc and without any sort of unifying work or principle. You’re not going to find much if any unifying work between social choice theory, therapy, and liberating structures.

This matters because it means we’re not taking the active development of social technology seriously. This shows in a couple of ways:

  1. There is very little cross-pollination between these sectors because nobody realises they’re doing the same things. RPG developers know a lot about how to manage unruly players sitting around a table, but this knowledge rarely makes it into meeting design.
  2. There is a huge amount of low hanging fruit available to be picked.
  3. Where we develop social technology we do not do it with the understanding that it is a technology (with the partial exception of the games sector).

In particular because we do not deliberately design social technology, most social technology is very poorly designed.

Have you noticed how many startups pivot? One major reason for that is that it’s really hard to build the right thing. When building new technology you will run into at least one and usually several of the following:

  • You misunderstood what the users wanted
  • You misunderstood what the users needed
  • You misunderstood what the users were willing to pay for
  • You failed to predict how people would use your technology
  • The solution you imagined doesn’t actually work
  • You ran into a hard problem that you couldn’t have anticipated

And many more. If you just build something and hand it to users, the results will rarely go very well. To build something that works you need to prototype, playtest, and iterate. Anything else is doomed to failure or, worse, mediocrity.

Almost none of our widely deployed social technology is designed in this way, and it shows. The best we can hope for is that because it’s easy to adapt and copied widely it’s subject to a certain amount of evolution over time, so the worst of the rough edges are knocked off and some of the best ideas are preserved, but often what happens instead is that we take something that doesn’t work very well and lock it in as a cultural tradition, with all its flaws intact.

Take my example of TickTalk. TickTalk comes from me taking Lean Coffee and playtesting it a bit, deciding it wasn’t very good4, fixing the bits I didn’t like and then playtesting it some more until we’d refined it down to a system that worked well.

Another thing that the lack of deliberate design means is that we often do not make good theory informed choices of social structures. Take for example my suggestions in Democracy for Lunch that you should use random ballot or approval voting for most group meal decisions. This is an example of providing a piece of social technology (a voting system) for a commonly encountered problem. This happens fairly rarely. To the degree that most people use social technology to decide on lunch options, it tends to be a mix of dictatorial (suboptimal: Does not involve whole group), plurality (the worst voting system) and consensus forming (comparatively high effort for a comparatively low importance option).

Because we don’t take the problems of social technology seriously, we end up in a situation where we often lack basic social technologies, or the ones we have and use are not very good.

Having an actual social technology sector would help fix this by making development of social technologies deliberate. The key benefits of this are:

  1. It turns the creation of social technology a problem that we can work at getting better at rather than something happens by accident.
  2. It would provide resources that are lacking for intensive play testing of new social technologies, providing us with more refined systems.

Why don’t we have a social technology sector?

Generally the key reason any given thing doesn’t exist is just that nobody has made it yet, but I think there are a couple of major factors that mean that it’s a bit tricky to develop such a sector.

The first is that it falls into an annoying middle ground area where it’s hard to make money out of, because once developed social technologies are cheap to copy and hard to prevent copying, and hard to do academic research on, because discovering universal truths about human nature is not an easy problem. As a result, it’s hard for there to be a sector per se, because industry and academia will both struggle to create it.

One way to make money out of it is basically management consultancy and writing books. The problem with doing this is that for the most part it doesn’t seem to matter if the social technology you’re developing actually works, because people will turn it into an ideology and make money off it either way. As a result you’re better off investing money into marketing than technology development even if you work with social technology5.

Academic research on the other hand struggles because running experiments on people is hard. A good example of this is the research on brain storming vs brain writing (the difference is basically how much you use speaking vs writing). You can come to some tentative conclusions, but really it should have just been a bunch of play testing sessions. These wouldn’t come to Truth in the same way, but that’s ok – we’re not really interested in truth, we’re interested in constructing technology that more or less works.

The second is that I think people are resistant to the very idea. Social technology feels like something that we shouldn’t need, and a bit non-serious. All getting together to obey a system of rules in aid of some goal feels a bit game like to people, aren’t we past that?

The opposite of that is that people take their social technologies that they do adopt very seriously indeed. People who pick up social technologies and turn them into religions, investing thoroughly in every aspect of the technology. Listen to someone telling you about Kanban, or Lean, or Democracy (which always means the specific version of Democracy in their home country of course). They often struggle to conceive of it being legitimate to do things any other way, which impairs the development of these technologies.

How do we get to a social technology sector?

If you gave me a couple million pounds to bootstrap it, I’d put together a cross disciplinary team of a couple people (say, a management consultant, a role playing game designer, a couple of academics) and start trying to build a business going into companies and developing social technologies with them to try to streamline their workflows. This is not particularly my current plan, but hey if you have a couple million pounds you want to send me way I wouldn’t say no.

The more accessible route and the one that I hope we adopt is to be the social technology sector you want to see in the world.

Social technology is cheap, adaptable, and easy to copy. This makes it extremely amenable to grass roots experimentation and open source development.

Step one of this is to let go of the idea of it being weird. We can start to be more deliberate about the social structures in our lives, and from there we can start to experiment with them, and when we do, we can write about them and share them with the world. Other people can then take our technologies and experiment and build their own.

If you’re wondering where to get started, my recommendation would be to find a couple of interested friends and start a structured discussion with them. Either TickTalk or Conversation Cafe are good starting formats.

You could even go meta while doing that – start thinking about other areas in your life where you could take more a deliberate, explicit, approach and ask what technologies might enable that.

This entry was posted in Uncategorized on by .

The Ethics of False Negatives in Interviewing

When interviewing people, there is one significant ethical component to the decision making process that is rarely made explicit, and that is often done badly as a result. I’ve written about it previously (here, here, and here), but you don’t need to read those – this is a standalone piece that sums up my current thinking.

The problem is this: The natural shape of the problem of interviewing encourages you to hire people who are similar to yourself. As a result, the aggregate effect of seemingly innocuous and effective decision making processes is often to amplify structural inequalities in your industry.

It will take a while to unpack how this plays out, so bear with me.

I’ll start with a simplification and only consider the case where you’re hiring one person for one role. Many real interview processes don’t play out this way – you might hire multiple people, there might be some flexibility about roles, etc. The dynamic is clearer in the one person to one role case, so I’ll stick to that, but most of the same considerations apply in general even if the details change.

When you are running such an interview process, your goal is to sift through a lot of candidates and hire one who would be good for the role (The question of what “good in that role” means is very complicated and could easily take up an entire post on its own. I intend to ignore it entirely, as others are much better placed to talk about it than I am). Effectively, you see a series of people, and make a yes/no decision about hiring each of them, stopping when you’ve hired enough people.

When making this yes or no decision to hire someone, there are two ways it can go wrong: You can say yes when you should have said no (a bad hire, or false positive) or you can say no when you should have said yes (a bad rejection, or false negative).

The most important thing to understand about interviewing is that you are extremely concerned with the false positive rate, and only somewhat interested in the false negative rate. That is, you want to ensure that the people you hire would be good in that role, and you are less concerned with the question of whether the people you reject would have been bad in that role.

Why? Well, because the cost of a bad hire is almost always much higher than the cost of keeping on trying unless your standards are incredibly stringent. You pay an opportunity cost, but as long as there are a lot of candidates, the opportunity cost of rejecting any one good candidate is low. All you are really paying is the cost of interviewing, so as long as the base rate of good candidates is tolerably high and your false negative rate isn’t too high, it’s mostly OK.

Why does it work out this way? It’s because the cost of a false positive is high and visible, while the cost of a false negative is low and invisible.

Because you have no further contact with most rejected candidates, a false negative and a true negative are functionally indistinguishable – if you could tell they were a false negative they wouldn’t be a false negative! As a result, if you have a high false negative rate it won’t look like a high false negative rate, it will look like you’re getting a lot of bad candidates and the cost of interviewing is correspondingly high.

In contrast, if you hire a candidate, you now have a lot more information about them. You will find this out over the coming months and years, and will eventually become reasonably certain as to whether your hire was good or not. This means that false positives will almost always eventually become visible to you, but they will do so at great cost! They’ll have spent a significant time as dead weight (or active toxicity) in your organisation, and this opportunity cost is large: You were hiring because you needed someone to help you out, and you’ve now spent months or years not getting that benefit.

As a result, every false positive is both extremely costly and eventually discovered. This means that you have both a strong incentive to keep it low, and good feedback that allows you to do so.

As a result you can roughly think of “rational” interview design as about minimizing the false positive rate while keeping the cost of hiring to some reasonable level.

Before I go on I want to emphasise that this is very reasonable behaviour. Asking a company to ignore or substantially raise its false positive rate is not going to go down well, and is more likely to result in a kind of theater of signalling where they find more complicated or worse ways to get the same benefit.

However, it’s worth thinking a bit more about the structure of this process, and to try to shift it a bit, by adding new constraints and incentives.

Why? Lets turn to a fact that we have been ignoring so far: Candidates are people, with their own needs and motivations.

A candidate is similarly paying a cost for interviewing, but their interests in the error rates are reversed. The cost of false negative rate is almost entirely paid by the candidates, because it means that they are denied something they wanted (and deserved), and there is a high emotional impact to rejection. With interviewing, they also have a significantly higher opportunity cost, because there are fewer companies they can reasonably apply to than you have candidates applying (this is less true in other examples of this dynamic).

As a result, false negatives are not nearly as free as they looked. Instead they are what economists call an externality – something that looks free because someone else, in this case the candidate, is paying for it.

How much should this matter to you as an interviewer? Well, some, certainly. If you want to behave ethically as an interviewer and as a company you do need to at least consider the harm to the candidate, but anecdotally it seems to be a thing that people are vaguely aware of and consider to be an acceptable cost – after all, most interviewers have been interviewees, so they have some idea of what it’s like to be on the receiving end. So for most companies there is already at least a certain amount of respect for the candidate’s time (large, self-important companies, with long multiday interview processes notwithstanding). It’s a thing that could be considered more, certainly, but it’s not a huge moral crisis.

Unfortunately “respect for the candidate’s time” does not fully capture the cost of false negatives, because not all false negatives are created equal.

We now need to unpack another thing that we’ve been ignoring so far: Interviewing is not a magic black box that spits out an answer of good or bad, it’s a reasoning process based on interacting with the candidate.

In an ideal world you would have a perfect simulation of what it would be like to work with the candidate, and hire them if that simulation was positive, but in the actual interviewing process where you have a small amount of time you basically just ask them some questions, get them to perform a task, and make your best judgement based on the evidence.

Again, this is fine, there’s literally nothing else you can do, so you should do that, but you shouldn’t do it uncritically, and it is worth thinking in more detail about the specifics of the reasoning process you have.

The core problem is that other interviewers are likely to reason in similar ways to you. This means that individual candidates may experience a much higher false negative rate than average.

Take, for example, someone who finds interviews very stressful and thus underperforms in them compared to their actual job ability. They will experience a significantly higher false negative rate than average, and experience a correspondingly higher cost to interviewing1.

When the variation is low, it’s tempting to not worry about this that much – so what if a candidate has to interview at twice as many places to get a job? That’s not your problem, and it doesn’t seem to be that big a deal.

Unfortunately there’s no reason to expect the variation to be low, and if some people find they are rejected vastly more often than average, those people are at a significant disadvantage in life.

When you participate in a system which significantly disadvantages people like this, you need to think long and hard about whether you are doing them an injustice. I think in this case we clearly are.

In the worst case you could imagine people who will never get hired even when they would be very good in the job. These people will experience a 100% false negative rate no matter how good your false negative rate is on average.

How might such groups arise? Well, racism for starters. In a society with pervasive and overt racism (e.g. apartheid South Africa, the USA during segregation) you might be entirely excluded from jobs simply because of your race. In a society where people pay lip service to not being racist, the numbers will look less extreme, but as long as there is significant prejudice against a group that group will find it harder to get hired.

There the corresponding ethical advice seems easy, right? Don’t be racist. Job done. Unfortunately, it’s not nearly so simple as that.

The problem is that you can get broadly similar effects without any active prejudice against your part. Suppose you had a test that gave you 100% accuracy on 80% of the population – that is, the test says you should hire the person if and only if you should actually hire the person – the remaining 20% of the population always failed the test. If you ignore the population effect, hiring based on this test looks very good. It has no false positives, a fairly low false negative rate, and permanently marginalizes a fifth of the population.

How plausible is it that such a test exists?

Well, that’s an extreme case, but when hiring for software developers I think a reasonable case can be made that looking for open source contributions is a decent approximation to it. It’s certainly not the case that all open source developers are good hires, but looking at someone’s open source code is a pretty good signal of whether they’re good at software development. However, this means that people who don’t contribute to open source get left out. If you use open source as a necessary signal, you’ll exclude a whole bunch of people, but even if you use it as a sufficient signal, it gives people who don’t contribute to open source a significant handicap, and open source contributions very disproportionately come from well off white men, so there’s a strong structural bias there.

This should be the point where I tell you the right thing to do here, but I honestly don’t know. I’d certainly say you shouldn’t require open source contributions, but I don’t actually think that it’s reasonable to say you should ignore them. Primarily because even if I told you that you’d ignore the advice, but also because that in itself would exclude a bunch of people who e.g. have no formal education and want to be able to use their open source contributions as signs of competence.

Even if you manage to rid your interview process of tests like this that are structurally prejudiced against a group, it will be hard to entirely remove biases. The problem is that fundamentally selecting for minimizing false positives gives an intrinsic advantage to people who you understand how to interview. These will disproportionately be people who are similar to you.

Suppose you are interviewing someone, and you want to get a sense of whether you’d like working with them. In order to do this you need to have a conversation. If they are someone who you share a lot of culture with, this is easy – you have things to talk about, you share a language (both in the sense that your first languages may be the same, but also just in the sense of shared cultural references, jokes, etc). It’s easy to talk to them.

In contrast, someone from a very different culture will take more work – you need to establish common ground and figure out a mode of conversation that works well for you. If you were working with this person you would get a long period of time to do that, but in an interview you have an extremely short time, and as a result the conversation will often flow less naturally than it might. As a result you are less able to tell whether you will actually get along well with this person when working with them, and this again gives the people who are similar to you an advantage.

This pattern repeats itself over and over again: If you are in familiar territory, you know how to predict accurately in that territory, so when you come to try to reduce false positives you will automatically select for the familiar – the unfamiliar is uncertain, and so your false positive rate goes up.

How do you get familiar with what working with someone from a particular group is like? Well, you work with them. Which you’re less likely to do if there’s some significant structural prejudice against them in the interview process. In this way, structural prejudices tend to reinforce themselves over time – we hire people who are like us, and this increasingly refines our models of which people are good to work with into ones that are based on that familiar set of people.

None of this is to deny that there aren’t myriad other structural prejudices in interviewing, this one is just important to highlight because of how insidious it is: It doesn’t operate through any negative belief about the group being disadvantaged, it acts solely through uncertainty and a reasonable set of of behaviour following incentives, and so even people who are genuinely committed to doing better can fall victim to it without noticing.

Naturally at this point I should tell you about the solution. Unfortunately, that’s mostly a hard pass from me, sorry. The general shape of the solution is “get better at interviewing”, and I’m not actually good enough at conducting an interview process to really offer advice on how to do so.

So, to the degree I have advice, it’s this: Most interview processes I’ve been a part of (on either side of the table) have been quite slapdash, and that’s only going to exacerbate this problem. Given the impact of the natural pressures of interviewing, this has to change. Interviewing is a hard problem, with huge social impacts, and it deserves to be treated as such.

As a result, I would like people to think much harder about designing their interview processes, and do some reading and learn how to do it properly.

Most of this has to be fixed at the company level, but if you have any good resources on e.g. reading, courses, etc. that people can take, please feel free to leave comments on the blog or let me know through some other medium.

This entry was posted in Uncategorized on by .

How to do hard things

This is a system I only somewhat tongue in cheek refer to as “The Fully General System For Learning To Do Hard Things”. It’s a useful conceptual framework for how to get better at things that you currently find difficult.

I don’t always explicitly follow it, but I often find that when I’ve succeeded my success comes from implicitly following it, and almost every time someone asks me for advice on learning to do things I just describe a specialised version of the system to them.

The system “always” works in the sense that “eventually” either you will find out why the objective is impossible for you or you will succeed, but it’s very much the unhelpful kind of eventually where there’s no guarantee that it won’t take an interminably long time. The more likely outcome is that either you will succeed relatively quickly or get bored and give up, but that’s OK – the system is designed so that you will have gained benefit from following it at every step along the way even if you do not achieve your final goal.

I should also note that this system is not in any way a short cut. It’s a lot of work. The goal of the system is not to save you work, it’s to ensure that the work you do is useful.

The Single-Loop System

When you know what success looks like but cannot currently achieve it, the system works as follows:

  1. Find something that is like the hard thing but is easy.
  2. Modify the easy thing so that it is like the hard thing in exactly one way that you find hard.
  3. Do the modified thing until it is no longer hard.
  4. If you get stuck, do one of the following:
    1. Go back to step 3 and pick a different way in which the problem is hard.
    2. Recursively apply the general system for learning to do hard things to the thing you’re stuck on.
    3. Go ask an expert or a rubber duck for advice.
    4. If you’re still stuck after trying the first three, it’s possible that you may have hit some sort of natural difficulty limit and may not be able to make progress.
  5. If the original hard thing is now easy, you’re done. If not, go back to step 2.

The reason this works much better than just practicing the hard thing is because it gives you a much more direct feedback loop. There is exactly one aspect of the problem at any time that you are trying to get better at, and you can focus on that aspect to the exclusion of all else. When you are practicing something that is difficult in multiple ways, you will be bad at it in all of those ways. More, you will be worse at it in all of those ways than you would be if you’d tried them on their own. Additionally, when you fail you have to do a complicated root cause analysis to figure out why.

Instead, by isolating one aspect of the problem that is difficult, you will fairly rapidly improve, or hit the limits of your ability.

The Double-Loop System

If you don’t know what success looks like, you need to do double loop learning, where you mix improving your understanding of the problem with your ability to execute the solution.

  1. Apply the single loop system to the problem of improving your understanding of the problem space (e.g. consume lots of examples and learn to distinguish good from bad) in order to acquire a sense of good taste.
  2. Apply the single loop system to the problem of doing well according to your own sense of good taste.
  3. Get feedback on the result from others. Do they think you did it well? If yes, great! You’re good at the thing. If no, either improve your sense of taste or theirs. If you choose yours, go back to step 1 with the new example. If you choose theirs, apply the single loop system to the hard problem of convincing others that your thing is good.

Is this all a horrible oversimplification? Well, yes, of course it is. It is however a very useful horrible oversimplification that is very good for getting you unstuck when problems seem intractable.

How To Identify Points of Difficulty

Sometimes it will be obvious what you need to improve, sometimes it won’t. When it doesn’t, here are some things that can help you figure it out:

  • Try to do the thing as best you can. Don’t worry about failing, failing is expected, but try to pay attention to how you’re doing it. Write down a list of things you thought you did badly, and things you did adequately but struggled with. Also if at some point you got stuck, note where you got stuck.
  • Look for beginners exercises for the area you want to work on and try a variety of those. Observe which ones are hard.
  • Talk to an expert on the subject (ideally one who is used to teaching) and ask them to help you identify some areas you need to work on.
  • Rather than starting from the easy thing, work in the other direction. Try taking the hard thing and removing one hard aspect of it at a time until you get it to a point where removing any hard aspect would make it easy.

Worked Example: Learning to Write Better

This is particularly good as a mechanism for improving your writing (and writing about it is a good lead in to the mechanism on a particular area, so I’d encourage everyone to work on improving their writing).

Writing can be hard in a wide variety of ways. Some common ones (in roughly the order I think it’s worth tackling them) are:

  • The actual physical process of writing.
  • Finding a good writing voice.
  • Perfectionism and/or fear of showing others your work and/or not being sure what to write about stopping you from the process of starting to write about it.
  • Editing and structure.

And that’s even before you get into specific forms of writing. You could also struggle with e.g. dialogue, description, etc.

Set against this is the fact that if you’re reading this you definitely can write. I promise. You might not be able to write a novel (I can’t at the moment), but you can certainly write a tweet, and it’s just a series of incremental steps to get from there to wherever you want to be.

Here’s some examples:

  1. Learn to touch type if you can’t already. If you can’t touch type you will get blocked on the basic mechanics of writing. This makes the feedback loop for everything harder. Maybe try ztype, typesy, or Mavis Beacon1
  2. Learn to write in your speaking voice. If you can’t figure out how to write about a topic, try speaking about it into your phone (get a recorder app). Once you’ve started writing, try to read it out loud. This isn’t the only writing voice worth using, but it’s an important one.
  3. Practice writing without any expectation of quality. e.g. write in a private file or google doc, or create a blog with an explicit disclaimer that it’s for writing experiments. Some good writing experiments:
    • Observational writing. Pick an object (a lit candle is an interesting one) and write down everything you notice about it.
    • Pick a subject you are reasonably familiar with and set yourself a word count of, say, a thousand words. You can write whatever you want as long as you hit that goal. Feel free to write a stream of consciousness.
    • Pick an opinion of yours and write a 500 word case for it.
  4. Rather than worrying about the general problem of editing, start thinking about editing with specific goals. e.g.
    • This ties in well with getting a good writing voice. Read it out loud, figure out where good pauses are and put paragraph breaks in where natural pauses occur. Fix language that sounds awkward.
    • Try editing purely based on word count. Can you express the same thing with fewer words? What would you do if you had to cut the length of the piece in half?
    • Try editing to remove specific words. Can you write the whole thing in up goer five? (Note: The result of doing so will be terrible, but the goal here is to practice editing more than produce a good piece)

There are plenty of other things to try, but these are some good starting points.

This entry was posted in Uncategorized on by .

The One Weird Word Rule

Here is a writing trick I find useful (though am not always good at following): You are only allowed one1 weird word2 per article, and you have to explain the meaning of that word.

If you have more than that, you’ll almost certainly lose the reader because they will not be able to keep track of the multiple new words.

What makes a word weird? Familiarity for the audience. If I’m writing for a monolingual French audience then they have my sympathies because my French is somewhere between appalling and non-existent and every English word is weird to them. Similarly, if I use a technical term, that term is weird for anyone outside that technical speciality.

If you wanted to ensure your audience is fully general (among English speakers) you could go full up goer five3 in the article, restricting yourself to only the thousand most common words in the English language. This is a useful exercise but I think it’s going a bit far.4. Use your judgement – it’s generally reasonable to assume that the reader is fairly literate, it’s generally not reasonable to assume they know what doxastic means5 6.

This is the real power of the trick: Because you cannot determine if a word is weird without knowing the audience, by following this rule you first have to decide who that audience is. This makes you think about the level of your writing, and what language it is appropriate for you to use and what you should avoid.

Once you’ve done that, the one weird word rule can be seen as a leniency rather than a restriction. It’s not that you’re only allowed one word your audience doesn’t know, it’s that it’s OK to use one word your audience doesn’t know. As long as you explain it well, a single article is about the right amount of text to get someone comfortable with a new word, and it’s fine to try to expand people’s vocabulary where it would be useful to do so, but if you try to do too many new words too fast your audience will quickly get lost, and your point will not come across well.

This entry was posted in Uncategorized on by .