The Google memo and freedom of speech, part 3: how not to make a controversial case.

This is the final part of my reflections on the controversy around the anti-diversity that got software engineer James Damore fired from Google. This part delves into the murky content of the infamous 10-page document at the heart of the affair.

The other two parts dealt with, firstly, with some of the problems that bedevil discourse around freedom of speech and, secondly, the political philosophy underpinning the concept (but you don't have to have read either to make sense of what follows). They tend to lead to the following conclusions:-

(A) Saying something controversial such as suggesting there might be biological differences between men and women should not automatically lead to one's firing. (There is a public interest in a reasonably free and open discussion of ideas). But equally

(B) The fact that such comments could be said to be in the interest of the free exchange of ideas should not automatically protect one from being fired for this. (You don't have a moral, let alone a legal, right to make political/ideological statements as part of your job.)

So what is an employer to do if a 10-page rumination on such a theme, or some similarly controversial topic, starts circulating the office? The following factors might be relevant:

• Does it appear to be good faith advocacy for a change that the employee would like to see implemented? Or is he or she just trying to ruffle feathers and cause a nuisance?
• Is it closely targeted at some particular policy or else is it a generalised attack on the company's values?
• Has every attempt been made, where controversial statements are made, to avoid causing offence (or worse, belittling others)?
• Is it politely written, or is it instead aggressive or insulting?

Let's now turn to what James Damore wrote and circulated.

Here's a link to the full memorandum.

Read an extended sample of a person's prose and you get a sense of the mental world that they inhabit. This is not always an edifying experience. Spending any time with the thoughts of James Damore (as I did to write this piece) is rather like being stuck in a lift with a spotty, know-it-all 10-year old: it's not just the lack of basic emotional intelligence (something that is anyway a badge of pride for people like him: witness his call in the paper to “de-emphasize empathy”) but the pervasive intellectual ignorance, the lack of familiarity with ideas originating from outside of his very narrow fields of interest.

To trot out the line, for example, “[w]e have extensive government and Google programs, fields of study, and legal and social norms to protect women, but when a man complains about a gender issue issue affecting men, he’s labelled as a misogynist and a whiner” without recognising how desperately hackneyed this form of argument is, how many millions of times it has patiently been dismantled (not just in the sphere of the New York Review of Books or whatever but in ordinary living rooms, pubs and secondary schools around the world), marks one out as the sort of person who has never had a serious discussion with another adult about politics or read a broadsheet newspaper let alone developed a moderately cultured awareness of the major currents of contemporary thought.

The problem with accusations of pervasive bias

The title – “Google's Ideological Echo Chamber” – sets the tone. Damore certainly didn't hold back in critiquing his employers. He accuses Google of harbouring “several biases” that have conspired to create a “politically correct monoculture that maintains its hold by shaming dissenters into silence”. They are allegedly “blind” to alternative points of view whilst promoting “encroaching extremist and authoritarian policies”. It's tin foil hat stuff from the get-go. The word “bias” (or “biases”) appears a total of 37 times.

What is objectionable about Damore's use of the notion of bias is more than just a matter of the unbalanced, aggressive tone in which the point is pursued – although this aspect is bad enough and isn't at all ameliorated by the document being peppered with nauseatingly patronising platitudes (“[t]hankfully, open and honest discussion with those who disagree can highlight our blind spots and help us grow, which is why I wrote this document”). It goes to the substance of his position in several ways.

For one thing, the across-the-board extreme generality of the accusation of bias calls into question the entire bona fides of Damore's complaints. If you are seeking persuade another of anything you clearly have to be careful when suggesting that that person may have overlooked a particular factor: one does not want to be insulting or call into question their intelligence or integrity. If, instead, you tell the person that you are imploring that they have not merely overlooked a particular point in coming to a particular conclusion but suffer from a pervasive, thorough blindness that means that they cannot safely trust their own judgement on anything what on earth are they supposed to with this information? Resign from whatever position they hold? As a litigation lawyer I know that it is never a good move to accuse a judge of bias unless you think this will result in his or her recusal. Because otherwise you just end up with a pissed-off judge deciding on your case; Emerson's maxim “when you strike at the king, you better kill him” applies. My objection here, to be clear, lies not in the mere fact that Damore's approach is offensive to its target – lots of legitimate political writing is precisely that – it is rather that his target is one and the same with the person he is ostensibly seeking to petition, which calls into the question the genuineness of the whole endeavour. Did he really think that Google would respond by saying “hey, yeah, we are fuckwits, aren't we”?

Moreover, it can hardly be Google's management alone that suffers from biases. As Damore himself recognises at certain points in the paper of course we all do. But that being the case, one has to assume that the validity of the positions that Damore himself is advocating may be undermined by an equally pervasive blindness on his part. Ultimately then any suspicion of pervasive blindness ought to drop out of the picture because it appears on both sides of the equation. Whining on about it is therefore at best irrelevant, at worst a tedious diversionary tactic.

Good, bad and neutral biases

Even more objectionably, albeit more subtly, Damore uses the notion of bias in a specious way to attempt to relieve him of the burden of engaging with any of the actual merits of arguments that run against the positions he advocates. This strategy is tied up with the fact that there are at least three, quite radically different meanings of the word “bias” that are implied at different times by Damore's text. Not only does he fail to explicitly distinguish between these different meanings, but, as I will show, his argument relies actively on their conflation.

1. Bias in a distributional sense. People sometime use the word more-or-less neutrally to describe the relative distribution of elements in some state of affairs, eg. “there is a bias in the production of jelly beans towards red ones”; “Tim Greenhat's investment portfolio is biased towards North American small company stocks”.

2. Cognitive biases. The concept of cognitive bias, pioneered by psychologists Daniel Kahneman and Amos Tversky, describes certain common reasoning processes that lead people to draw illogical inferences or form objectively irrational conclusions in certain circumstances. In simple terms a cognitive bias is a tendency to make certain kinds of mistakes. For example, confirmation bias – which is name-checked in Damore's document – describes the human tendency, confirmed by lots of psychological experiments, to interpret new information in a way that confirms one's preexisting conceptions, including by discounting conflicting information. Cognitive biases are a trendy topic at the moment, with Kahneman (who won the Nobel prize for economics 15 years ago, study of cognitive biases being fundamental in behavioural economics) publishing a bestselling book on the subject a few years back (which is well worth reading). Although some cognitive biases might plausibly be evolutionary adaptive (such as the self-serving bias: the tendency to claim more responsibility for successes than failures) they are generally thought of as bad things. At least they always involve mistakes in the sense that the person suffering from such a bias is attempting to complete a specific mental task (working out which thing is most likely to happen, or determining which course of action will be most profitable, etc) but fails to achieve this.

3. Moral or ideological bias. Bias can also be used to describe our basic preferences, including our moral preferences. In his essay, Damore treats “preference” as identical to “bias”. (Some might think such usage is rather idiosyncratic but I will not quibble on that ground alone.) Hence “compassion for the weak” is a bias (a “Left Bias” according to Damore). It is a bias in a neutral sense because faced with a choice between two options, one of which favours the weak and one of which does not, the person with this preference all other things being equal will select the former. But by this logic there is no need to stop at the preferences that divide us politically. We should also include in the list of “biases” the various preferences that most humans share, for example:

• The aversion to using force or fraud to acquire resources
• The preference for protecting rather than harming children
• The tendency to praise people who act altruistically and conversely to criticise those who act selfishly
• The preference for making profits rather than losses
• The preference for happiness over misery as regards oneself and others

Such biases are essentially the entire content of human motivation and morality. They are very often, albeit not always, regarded as good rather than bad things.

Although not specifically invoked by Damore, there is also a fourth meaning of bias:

4. Bias as a lack of impartiality. I mentioned above the practice applying for a judge to recuse herself for bias. That kind of bias – a lack of impartiality, due to, eg some financial tie to one of the parties – only arises when somebody has a pre-existing duty to act fairly between competing parties because they are performing a specific role such as judging a lawsuit. This kind of bias is definitely a bad thing, although it's really not an everyday concern (unless you happen to be judge).

What Damore does is conflate the first three types of bias whilst trading on the negative connotations of the 2^nd as well as, I imagine, the 4^th in order to belittle any attitude that falls within the 3^rd category, ie moral and political preferences, where this conflicts with his own.

This is childish and absurd. For one thing, category 3 “biases”, as mentioned above, form essentially the whole of morality (as Damore kind of acknowledges: “Political orientation is actually a result of deep moral preferences and thus biases.”) You can label moral preferences “biases” if you want but if you do you cannot legitimately trade on the negative connotations of the word “bias”. It is a good thing that people favour those who help others. It is a good thing that people abhor violence and harm to children. It is a good thing (I think, at least) where people show compassion to the weak.

One cannot conflate moral/political preferences with cognitive biases because the former are preferences whereas the later are simply tendencies to make mistakes. To take a political preference that I do not myself share, namely “respect for authority” (correctly placed on the right-hand side of Damore's helpful ready reckoner on page 2 of the memo), whatever arguments I might legitimately use to range against maintaining such a preference, I cannot simply condemn it as a mistake, as a mere deviation from some other attitude that is “true” because it sits in the middle of some distribution curve of attitudes that various people hold. A preference for authority cannot be simply mistaken in the way that a judgement, assessment or calculation can be because, being a preference, the person holding it considers it to be a valid motivation. The fact that it leads to a particular weighted pattern of outcomes – bias in the 1^st sense above – can be no criticism: that is the whole point of a preference. It will either be a fundamental preference, ie one that cannot really be justified beyond the fact that it is preference that one holds (love is a common example), or else it is a preference held for particular reasons, in which case I need direct my arguments to those underlying reasons rather than the "bias" caused by the resulting preference itself.

Not only is is faintly depraved to treat all moral preferences (which, again, include wanting babies not to be injured) as equivalent to the kind of mental slipshodery that Tversky and Kahneman identified as leading people to make bad business decisions, it also permits Damore to suppose he can completely ignore the reasons that often sit behind such preferences. It allows him, in short, to put any goal that he does not share in a kind of black box and ignore any sensible rationale that might exist in its favour – to instead treat it merely as a fault to be corrected.

Diversity

To take the central policy goal discussed in the document – that of increasing representation of women and ethnic minorities in tech – there are all sorts of possible underlying rationales for this objective.

One of these reasons actually has nothing to do with long-term social engineering (or social responsibility) and everything to do with directly getting the best people for the job. It goes under the name of diversity. Although Damore uses the word six or seven times throughout the document no-where does he demonstrate that he actually understands what the argument for it actually involves and why it might in fact justify what he calls “discriminatory practices”.

Diversity – having people of different backgrounds and identities – is regarded as a good thing, for one thing, because it broadens the pool of experience that a company can draw upon. It is possibly also broadens – particular if Damore is right that there are relevant cognitive differences between men and women – the pool of skills at the organisations disposal. It is also a good thing if, by making the workplace a more comfortable, less exclusionary place to be for a larger proportion of the population it increases the company's hiring reach.

I do not comment here on the validity or the strength of those claims. But if diversity is a good – and Damore says (genuinely or otherwise) that he agrees that it is – then the only non-trivial (ie not by accident) means of achieving that good will all involve, one way or another, deliberately engineering one's recruitment programme to achieve the hiring of a cohort of employees with a somewhat different gender/ethnic make-up to that which would be achieved had diversity not been a goal at all. In particular, believing that diversity is good necessarily means believing that the best person for the job is not going to be fully independent of the characteristics of the other individuals you have already hired under the same job description.

These points should not be vastly controversial to anyone familiar with basic investment principles. Suppose I run a pension fund and I have $10 million to invest. I do some research and I come to the conclusion that Stock A is likely to produce the greatest return, Stock B will produce the next best return, and Stock C will produce the worst return of the three. Should I put all the money in Stock A, since I think this is the most promising? Almost certainly not: it is generally better, the extent depending on your (or your client's) appetite for risk, to diversify one portfolio and hold a variety of different types of stocks.

One way of conceptualising this is to say that the prudent investor is effectively hedging against the possibility that application of her own preferred criteria for investment are fallible – that they will not, in fact, lead to the best outcome. On her preferred, primary criteria Stock A is the best bet. But the investor know from experience that her crystal ball is of limited power and often the 'best bet' loses. So she effectively suspends her own preferred primary criteria for selection by using a secondary set of criteria, namely those that ensure a diverse portfolio is selected.

Can it not be said that the employer who employs a hiring practices that – shock-horror – “effectively lower the bar for 'diversity candidates'” (per Damore) is doing essentially the same thing, namely showing a prudent scepticism about the objective strength of its own primary selection criteria? Damore does not even consider the possibility. The case for diversity with recruits is even stronger than with stocks because stocks do not, as employees do, have to pool skills to work effectively with one another or attract other stocks to apply for a job. Again, Damore ignores these points. It seems that to his mind the only true, objective criteria to select for whatever job it was that he had at Google were the precise ones upon which he himself happened to be hired.

Girls vs. boys

Finally, onto the most controversial aspect of the memo: the suggestion that differences between the representation of men and women in STEM fields is down to innate biological differences between the genders.

Given that Damore spends so much of his 10-pages of infamy whining on about other people's biases, as I've said above, to no actual purpose, it is ironic that his reasoning on this crucial aspect provides an almost a classic example of one of Daniel Kahneman's favourite cognitive biases: attribute substitution. Attribute substitution occurs when your brain is called upon to make some judgement that will be computationally complex to carry out and it cheats by substituting a slightly different question which is easier to assess.

The question that Damore claims to be answering is: are biological differences the cause of the gender gap in tech? What he actually does is answer the simpler but ultimately irrelevant question: are there some biological differences between men and women?

He starts by trotting off a bullet-point list of alleged personality differences between men and women (some of which are partially supported by proper research citations, some of which aren't). Women are more interested in people, on average, than men. Men like taking physical risks more than women do on average. That sort of thing.

But when he is done with his list, he carries straight on as if he has now fully explained the cause of the gender gap: on page 5, under the heading “Non-discriminatory ways to reduce the gender gap” he says “Below I'll go over some of the differences in distribution of traits between men and women that I outlined in the previous section and suggest ways to address them to increase women's representation in tech without resorting to discrimination.”

Whoah! Hold on. Where was the bit where Damore cleverly explains how men's incrementally higher scores on some or other spatial reasoning test or risk-taking metric means that one should expect the gender ratio in Google tech roles to be more or less exactly the 80:20 split that it is? Where was the bit where he carefully examined at all the other myriad seemingly very plausible contributing causes that have been suggested for the gender gap and showed them conversely to have negligible numerical impact?

These crucial parts of the essay of course don't exist. They don't exist because providing convincing arguments for or against (let alone proving) causation is fiendishly difficult in this context. Damore basically doesn't bother. Instead, he is content to rubbish the concerns of anyone who cares about gender inequality in the industry based on his own unproven pet theories.

Disambiguating the effect of social conditions is important. For even where purely biological factors stand at the beginning of a chain of causation that ultimately leads to a lot less women going into tech jobs that leaves wide open the question whether those biological factors operate through the intermediation of social conditions.

Think about it. Suppose I showed you evidence that there was a particular genetic marker that some people in a particular population had and that others did not and that the majority of people with that marker were illiterate and the majority of the people who lacked it were not. You'd agree that the cause of the gene-carriers' illiteracy was down to genetics, right?

But the situation I have just described would pertain precisely if you were comparing the literacy of black and white Americans circa 1850, during slavery when, of course, education was not generally extended to the slave population. Having the genetic make-up of an African-American would predict illiteracy; having that of a White American would predict literacy. Yes, having African-American genes was then causal of illiteracy. But, for obvious reasons, it was causal in an entirely trivial way.

Interests vs. abilities

In fact, one piece written in the aftermath of the affair that does make a fairly convincing case that biological factors make a somewhat less trivial difference to the tech gender gap is this one by Scott Alexander, which unlike Damore's screed is carefully reasoned and worth reading. It argues that the evidence points to hormonal differences between males and females affecting levels of orientation towards pro-social behaviour (ie helping people). This, it is argued, leads women, on average, to be more interested in activities and career paths that are people-orientated (teaching, medicine, etc) than are men.

Alexander's conclusions are controversial (a nice thing about his piece is that includes, at the end, a detailed response from the writer who Alexander is responding to, which gives a full-circle view of the scientific controversy). But even if we assume his thesis is correct it hardly supports the sort of sweeping conclusion drawn by Damore to the effect that efforts to encourage greater participation of women in tech are worthless.

For Alexander's analysis is that differences in interest, rather than ability, drive the tech gender gap. The logic of this position is that some lower-ability men must be displacing higher-ability women in coding jobs because those higher ability women are peeling away from the relevant career path to do other, more people-orientated stuff like teaching (or, equally, because men, with their weaker pro-social instincts, are failing to do so). Let me say this again: the logic is lower ability people displacing higher ability people. All because of some dumb hormones. Surely the existence of such a dynamic ought to concern us. Surely it ought to particularly upset Damore given his jobs-ought-to-be-filled-with-the-highest-ability-candidates-no-matter-what-the-cost approach. One would think, indeed, that even were it to prove fairly expensive and difficult to overcome such instinctive motivational differences we might want to make the necessary investment.

In fact, there is good reason to believe that it is not all that difficult to overcome such biologically based motivational factors. Steven Pinker makes the point in Better Angels that people (men in particular!) are born with instinctive biases towards aggressive behaviour that in civilised society we do all sorts of things (create and hold to codes of manners, scold our young, etc) to repress. The civilised habits that such social controls instill become, Pinker points out, second nature to us – in contradistinction to our first nature, ie the Lord of the Flies-type barbarism that we would sink into if we were all left to our biological instincts alone.

If we can devise customs, values and programmes of education which constrain successfully (read Better Angels if you have a problem with “successfully”) the natural violent tendencies of men, then it shouldn't be beyond our wit to make a few tweaks to society so that high ability potential female software engineers are coaxed into using their talents to the fullest and ignoring the genetic siren-call towards people-orientated professions (and vice-versa with technically more mediocre men).

Parting with Damore

None of the above considerations got so much as a look in with Damore's manifesto.

He took the view, instead, that flinging together some homilies about gender differences of completely indeterminate quantitative standing and sprinkling the lot with random citations was enough to give him a licence to go off on an Alf Garnett-style rant about political correctness prefaced by a slimy, philosophical incoherent invocation of the right to freedom of expression as a pre-emptive shield against the criticism that he was aware was coming his way right from the start. It wasn't. There was always too much at stake: the sense of empathy and solidarity to which his female co-workers were entitled, his employers' reputation, and ultimately his own job.

I don't buy the line that Damore's sacking will inhibit open debate on the issue. It is evidently possible to make sane observations as to how biological differences between people might affect how we ought to behave to one another, how our companies and institutions operate or how we formulate education policy. The writers mentioned above – Pinker and Alexander – and many others have done so and avoided censure. Why? Because, quite simply, they avoided doing so in a dickish way.

The central reason why Damore deserved to be sacked was because revealed himself in those 10-pages of bile thinly-disguised as a helpful comment-box suggestion to be an unpleasant piece of work. The intellectual dishonesty of the work, the dismissive treatment of weighty considerations counter to his position, and the shockingly arrogant tone would be repulsive enough were he simply asking for a raise or, say, a transfer to another office. But given all of this shit was being leveled at dismantling policies designed at assisting marginalised groups of people become less marginalised these things were unforgivable. Google said, in effect, sorry, but he just wasn't their sort of person, and they were right, I think, to do so.