Tuesday, May 28, 2019

The Problem of Self-Esteem: Comparison and Competition

Most of us want to feel good about ourselves. We want to think that what we do is worthwhile. We want others think well of our efforts. Some people claim to be indifferent to the opinions of others, but I think we should be sceptical of their claims. Their indifference is often an act — something they do, paradoxically, to attract attention and good will. They want other people to look at them and say ‘I envy their lofty indifference.’ Even if it is true that some people are genuinely indifferent to the opinions of others, I suspect they still want to feel good about what they do themselves, i.e. even if they don’t care about gaining the esteem of others, they care about gaining their own self-esteem.

But self-esteem can be a tricky thing to cultivate, particularly in a world of competitive endeavours. The philosopher Robert Nozick once argued that self-esteem was necessarily competitive and comparative: one person could gain self-esteem only if another person lost out. He also argued that the best way to overcome the injustice this might entail was to create a society in which there are many different ways to win self-esteem.

In what follows, I want to look at Nozick’s arguments in a bit more detail. I do so through the lens of Andrew Mason’s short article “Nozick on Self-Esteem”. As we shall see, Mason challenges some of Nozick’s foundational assumptions about the nature of self-esteem.

1. Is Self-Esteem Necessarily Comparative and Competitive?
To understand where Nozick is coming from, we need to understand self-esteem. Self-esteem is a positive self-assessment. It is a belief that you score highly on some valuable attribute. For example, you might attach your self-worth to skill in writing or artistic expression. In that case, you will have high self-esteem if you believe that you score highly on those skills.

But how do you know that you score highly on those skills? This is the critical question. Nozick’s claim is that the only way you can tell whether you score highly is by looking at how your performance ranks relative to that of other people. I will only know that I am a good writer by comparing what I write with other writers. Do I sell more books than they do? Do more people read my work? Do more top critics shower my prose with praise? Ditto if I am an artist. My assessment of myself is always conducted on a comparativist basis: am I better than others who do the same thing?

If Nozick is right about this, then it has some unpleasant consequences. It means that self-esteem is a necessarily competitive game. I only gain self-esteem if I am better than someone else at a particular skill. And my gain in ranking is their loss in ranking. Esteem rankings are thus a scarce resource: something that people fight over in order to gain self worth.

But is Nozick right about this? There is undoubtedly some truth to what he has to say. Sometimes the only way to tell whether you are good at something is by comparing your performance with others; and there is undoubtedly a lot of competition and comparison in the contemporary economy of esteem. But it is not the full picture. There are two reasons to think that Nozick’s endless war of all against all is avoidable.

The first reason is that some activities have objective standards of success: you can tell if you are good at them by comparing your performance to the objective standard. You don’t have to compare it with other people. Mason, working off an example provided by the philosopher Anthony Skillen, uses boat-building to illustrate this point. If you want to know if you are any good at boat-building, you see whether the boat you build floats, whether it gets from A to B and so forth. You don’t have to compare your boat-building with that of other people.

It’s not that simple and straightforward, of course. Even in the case of boat-building, you could, if you liked, start comparing your boat with the boats built by others. You might do this if you have attained the minimum threshold of success in boat-building (the boat floats!) and start to focus on other attributes of a boat such as its aesthetic beauty. You might do this to further hone your skills. But even then there will be a blending of objective and comparative standards at play. The important point, and the one that needs to be emphasised, is that Nozick’s mental model of self-esteem draws too much from the world of competitive sports, and this skews his reasoning into thinking that comparative assessments of worth are the only game in town.

The second reason for doubting Nozick’s argument is that many people accept that they will not score highly, relative to others, on some skills and attributes. Nevertheless, they can take pride in performing those skills to the best of their own ability. For example, I know I will never be as good a golfer as Tiger Woods, but I can take pride in playing the best round of golf that I am capable of playing. This optimisation of my own talents and abilities can be a source of self-esteem, even if my ranking compared to others is quite low.

Some people might argue that even this is implicitly comparative. In determining whether I have done something to the best of my ability then I am, implicitly, comparing myself to someone else who has not done something to the best of their ability (e.g. another golfer of a similar level of ability, or myself at an earlier point in time). But this can’t be true. As Mason points out:

…if it were true that the only way in which one could judge whether a person had done something to the best of their ability was by comparing their performance with others, then there would be grave difficulties in determining whether a person had done as well as they could. How could one tell whether a person had developed their talents to a greater extent than others, or had merely a greater initial endowment of them? 
(Mason 1990, 93)

The bottom line then is that although self-esteem is often comparative and competitive; it doesn’t have to be. Indeed, ideally, self-esteem should derive from a rational self-reflection on one’s own abilities, relative to absolute standards of success. This doesn’t have to be immune to the opinions and attitudes of others, but it doesn’t have to be hostage to them either.

2. The Distribution of Esteem
Even if we accept Mason’s critique, there are problems. He could be right, and yet it could still be the case that (a) many people continue to derive esteem from comparative rankings with others (i.e. they get trapped in the competitive ethos) and (b) they struggle to gain self-esteem even when following this approach because they cannot win the comparative ranking. What do we say to them? Can we help them to gain self-esteem? There are two things worth considering here.

First, it is worth considering Nozick’s own argument about how to deal with people who lose out in competitions for self-esteem: create more competitions. The tendency to measure one’s self worth relative to the skills of others is going to be a particularly problematic in a world in which there is only a handful of ‘games’ one can play to gain self esteem. To illustrate, imagine being a man in Ancient Sparta. Let’s suppose that in that culture the primary metric of self worth is courage and skill in battle (I’m sure this does a disservice to the historical reality but it conforms to the stereotype). If you aren’t a brave and skilful warrior, then you will probably lack self esteem. This could be true despite the fact that you have many impressive skills that can distinguish you from other people. Perhaps you are skilled dancer or artist for example. Unfortunately for you, there are no opportunities to compare your skills in those domains against others because no one else is being incentivised to participate in dancing or art competitions. If they were, you might find a way to gain self esteem.

This is why Nozick recommends that we create a society in which there is a diversity of competitions for self-esteem. This will allow individuals to find their own niche. I should say I use the term ‘competitions’ somewhat loosely here. The idea is not that there are lots of formal competitions taking place; rather it is that the society celebrates many different skills and allows for lots of informal comparative rankings to take place. As Nozick puts it:

The most promising ways for a society to avoid widespread differences in self-esteem would be to have no common weighting of dimensions; instead it would have a diversity of different lists of dimensions and weightings. This would enhance each person’s chance of…[making] a non-idiosyncratic favorable estimate of himself. 
(Nozick 1974, 245-246)

Now you could argue that this solution misses the point because we shouldn’t be deriving our self-esteem from these comparative and competitive rankings. We should try to shift to more objective assessments of ability. But that may not be easy in all cases, and I think there is something to what Nozick is arguing. A society that celebrates and tolerates many ‘competitions’ for self-esteem is going to be happier than one that channels everything through a few competitions.

This brings me to the second point that is worth bearing in mind. Nozick might argue that a free market, capitalist society is the best one for enabling lots of competitions for self-esteem. There are lots of jobs one can perform that are valued by the market, so you have an opportunity to find your own niche and gain self-esteem. But as Mason points out, one major problem in our society is that there is differential access to these different niches. Not everyone gets into the competition that suits them best. They are forced, through economic necessity, into jobs that do not suit their skills and aptitudes, but are a means to an end. Furthermore, the market doesn’t value all skills equally. Indeed, some skills have no economic value at all. People aren’t incentivised to develop those skills because of this. The net result is that many people end up in competitions that slowly gnaw away at their self-esteem.

This problem of differential access to esteem-raising activities has been a major theme of my own recent research on workplace automation (although I never talk directly about ‘self-esteem’, preferring instead to focus on meaning and well-being). I have argued that workplace automation can exacerbate the problems alluded to in the previous paragraph, but also provide the means to escape them. In other words, I have argued that Nozick’s ideal society in which everyone can find their own niche is more likely to arise in a post-work economy than in one committed to the capitalist work ethic. Defending this post-work ideal is a major focus of my forthcoming book Automation and Utopia.

Saturday, May 25, 2019

Discrimination and Fairness in the Design of Robots

[Note: This is (roughly) the text of a talk I delivered at the bias-sensitization workshop at the IEEE International Conference on Robotics and Automation in Montreal, Canada on the 24th May 2019. The workshop was organised by Martim Brandão and Masoumeh (Iran) Mansouri. My thanks to both for inviting me to participate - more details here]

I never quite know how to pitch talks of this kind. My tendency is to work with the assumption that everyone is pretty clever, but they may not know anything about what I am talking about. I do this from painful personal experience: I've sat through many talks at conferences like this where I get frustrated because the speaker assumes I know more than I do. I'm sorry if this comes across patronising to some of you; but I'm hoping it will make the talk more useful to more of you.

So, anyway, I am going to talk about discrimination and robotics. More specifically, I am going to talk about the philosophical and legal aspects of discrimination and how they might have some bearing on the design of robots.

Before I get started I want to explain how I approach this problem. I am neither a roboticist nor a computer scientist; I am a philosopher and ethicist. I believe that there are three perspectives from which one can approach the problem of discrimination and fairness in the design and operation of robots. These are illustrated in the diagram below.

The diagram, as you can see, illustrates three kinds of relationships that humans can have with robots. The first, which we can call the 'design relationship', concerns the relationship that the original designers have with the robot they create. Discrimination becomes a worry here because it might leak into that design process and have some effect on how the robot looks and operates. The second relationship, which we can call the 'decision relationship', concerns the decisions the robot makes with respect to its human users. Discrimination becomes a worry here because the robot might express discriminatory attitudes toward those users or unfairly treat users from different groups. The third relationship, which we can call the 'reaction-relationship', concerns the reactions that human users have to the behaviour of the robot. Discrimination becomes a worry here if the humans discriminate against the robot or if they learn and normalise such attitudes from their interactions with the robot and carry them over to humans.

A comprehensive analysis of the problem of discrimination in the design of robots would have to factor in all three of these relationships. I do not have time for a comprehensive analysis in today's talk so, instead, I'm going to focus on the second relationship only. That said, unless the robot is itself a fully autonomous agent, focusing on the second relationship inevitably entails focusing on the first relationship too since the robot's decision algorithms will be created by a team of designers. There is, however, a difference between them. Whereas the first relationship concerns both the look, appearance and general behaviour of the robot; the second relationship is concerned specifically with its decision-making practices and how they might affect human users.

With that caveat out of the way, I want to do three things in the remainder of this talk:

  • (a) I want give a quick overview of how philosophers think about the concepts of fairness and discrimination.
  • (b) I want to look at the debate about algorithmic discrimination/fairness and consider some of the key lessons that have been learned in that debate.
  • (c) I want to make two specific arguments about how we should think about the problem of discrimination in social robotics.

1. A Brief Primer on Fairness and Discrimination
Let's start with the philosophical overview and consider the nature of fairness. Fairness is a property of how social goods (i.e. money, food, jobs, opportunities) get distributed among the members of a population. A common intuition is that a fair distribution is an equal one. But what does that really mean?

To think about this more clearly, it will help if we have a simple model scenario in mind. Consider the image below. It represents a highly stylised social system. On the bottom of the image we have a population of individuals. These individuals are divided into three social groups (you can think of these as 'races' or identities, if you like). In the middle of the image we have what I am somewhat awkwardly calling 'outcome makers'. These are properties attaching to the members of the population that make them more likely to achieve certain socially desirable outcomes. These properties can take many forms. Some might be innate characteristics of the individuals (e.g. race, sex) and some might be more contingent or acquired properties (e.g. income, a good education, good health and nutrition). All that matters, is that they make it more likely that the individuals will achieve relevant outcomes. As you can see from the image, different individuals have different outcome makers and they are not evenly distributed across the population. Finally, on the top of the image, we have the outcomes themselves, i.e. the 'buckets' where the individuals in the population end up. For illustrative purposes, I've imagined that the outcomes are jobs but they could be anything at all (e.g. income, number of friends, access to credit and housing, number of intimate relations, whatever it is you care about). As you can see from the image, different proportions of the three main social groups have ended up in different outcome buckets. In fact, there is something oddly skewed about the outcomes since all the 'blue' members of the population end up in one bucket.

With this simple model in mind, we can explain more clearly some of the different ways in which philosophers think about fairness and equality .

Equality of Outcome: We can start with the concept of "equality of outcome", which is widely touted as a desirable goal for social policies. Following our model, this could mean one of two things. It could mean, in the extreme case, that all members of the population, irrespective of their social group, share the same outcome (in this case, they all have the same job but it could also mean they all have the same number of friends or income or whatever). This understanding is extreme and counterintuitive, at least in the case of jobs -- why would you want to live in a society in which everyone had the exact same job? -- so an alternative interpretation, which is more plausible, is that equality of outcome arises when all social groups are equally or proportionally represented in the different social outcomes. This corresponds to what some people call a principle of fair representation.

Equality of Opportunity: Even though equality of outcome is a popular idea, it is also widely criticised. People worry about a society that forces people into different outcomes in the interests of fairness. So instead of achieving equality of outcome they think we should focus on equality of opportunity. This is function of how the 'outcome makers' get distributed among the population. From our model, equality of opportunity would arise when each member of the population, irrespective of social grouping, is given a mix of outcome makers that enables them to achieve any of the different possible outcomes. This doesn't mean that they all have the exact same mix of outcome makers; it just means that whatever mix they have is such that they each have the same opportunity of achieving the different possible outcomes (the playing field has been levelled between them).

Theories of equality of opportunity are often complicated by the fact that different philosophers take different attitudes toward different outcome makers. A common assumption is that you cannot and should not equalise all outcome makers. For example, you cannot make all people have the same level of physical strength or general intelligence. Nor should you force people to acquire abilities that they don't really want (e.g. forcing everyone to take high-level quantum physics). You have to respect people's autonomy and responsibility for choosing their own path in life. This means that when thinking about equality of opportunity, you should equalise with respect certain kinds of outcome maker, but not all.

[A brief aside: you may notice from this discussion that I don't think much of the distinction between equality of outcome and equality of opportunity. My view is that opportunities are really just outcomes of a particular kind: they are outcomes that are steps on the road to other outcomes. But it would take a bit longer to justify this position, and the distinction between equality of outcome and equality of opportunity is popular one so I am working with it.]

This brings us to our second key topic -- discrimination. To understand how philosophers think about discrimination, we just need to add some details to our model. First, we need to think about how the members of the population access the different possible outcomes. I've assumed that this just a function of the outcome makers they possess, but that's not very realistic. In any real society, there will probably be some set of actors or institutions that decide who gets to access the different outcomes. We can call these actors or institutions 'the gatekeepers'. They act as screeners and sorters, taking members of the population and assigning them to different outcomes. To make it more concrete, and to continue with our example, we can imagine people interviewing candidates for different jobs and deciding who should be assigned to which job. Discrimination is a phenomenon that arises from this gatekeeping function. More precisely, it arises when gatekeepers rely on criteria that we deem to be unjust or unfair in screening and sorting people into different outcomes.

To understand this problem more clearly, we need to add a second complication to the model. This complication concerns the properties of the members of the population who get sorted into the different outcomes. Each member of the population will be a bundle of different characteristics and properties. Some of these characteristics will be 'protected' (e.g. race, age, religion, gender) and others will not be (e.g. income, educational level, IQ). The core idea in discrimination theory and practice is that gatekeepers should not use protected characteristics to sort people into different outcomes. They should only rely on unprotected characteristics.

Actually, it's a bit more complicated than that and we need to introduce several conceptual distinctions in order to think clearly about discrimination. They are:

Direct Discrimination: This arises when gatekeepers explicitly use protected characteristics to guide their decision-making, e.g. an interviewer explicitly refuses to hire women for a job.

Indirect Discrimination: This arises when, even though gatekeepers do not explicitly use protected characteristics to guide their decision-making, they rely on other characteristics (proxies) that have the effect of sorting people according to their protected characteristics, e.g. an interviewer refuses to hire someone with more than one career break (could be a problem if women are known to be more likely to have taken a career break).

Individual Discrimination: This arises when individual gatekeepers act in discriminatory ways (be they direct or indirect).

Structural Discrimination: This arises when social institutions, as opposed to individual gatekeepers, work in such a way that members of some social groups are systematically discriminated against when compared to others. Structural discrimination could arise with or without individual discrimination.

Positive Discrimination: This arises when gatekeepers are incentivised to use protected characteristics in decision-making in order to achieve a fairer representation of different social groups across the possible outcomes. This is usually done to correct for historic unfairness in social sorting (e.g. affirmative action hiring policies).

Impartiality: This is when gatekeepers show no favourability or bias toward certain social groups or individuals in their decision-making. This is often the long-term aim of anti-discrimination policies.

I appreciate that is a lot of conceptual distinctions but they are all important when it comes to understanding the debate about fairness and discrimination.

You might ask: "How do we prove that discrimination has occurred?" That is a good question and it is often difficult. Sometimes we have clear and unambiguous evidence of discriminatory intent, but more often we see that different social groups have been sorted disproportionately into different outcomes and we infer from this that some discrimination might have occurred. A more thorough investigation might confirm this suspicion. Another question you might ask is "how do we decide what counts as a protected characteristic?" This is also a good question and there is no single answer. Different moral considerations apply in different cases. Sometimes we designate something to be a protected characteristic because we believe it has no actual bearing on whether someone would be a good fit for a particular outcome, but people mistakenly think that it does, and we want to stop this from influencing their decision-making; other times it is because we don't want to punish people for characteristics that are outside of their control; sometimes its a combination of factors. There is an interesting phenomenon nowadays of something we might call 'protected characteristic creep', which is the tendency to think that more and more characteristics deserve to be protected against discriminatory decision-making, which often has the net effect of making it more difficult to avoid discrimination.

2. Lessons from the Algorithmic Fairness Debate
With that overview of the philosophy of fairness and discrimination out of the way let's consider the implications for robotics. And let's start by considering the lessons we can learn from the algorithmic fairness debate. As some of you will know, algorithmic decision processes have been used for some time in the public and private sector, for example, in credit scoring, tax auditing, and recidivism risk scoring. This usage has been growing in recent years due to advances in machine learning and big data. This has generated an extensive debate about algorithmic fairness and discrimination. Looking to that debate, is an obvious starting point for anyone who cares about fairness and discrimination in robotics. After all, the decision-making algorithms used by robots are likely to be based on the same underlying technology.

Some of the lessons learned from this debate are important but relatively unsurprising. For example, it is now very clear that decision algorithms can work in biased and discriminatory ways. This may be because they were designed to rely on discriminatory criteria (directly or indirectly) when making decisions, or it may be because they were trained on biased or skewed datasets. Trying to recognise and correct for this problem is an important practical concern. But, as I say, it is relatively unsurprising. I want to focus on two lessons from the algorithmic fairness debate that I think are more surprising and still practically important.

The first lesson is that, except in very rare circumstances, there is no way to design an algorithmic decision process that is perfectly fair and non-discriminatory.

This is a lesson that was first learned by investigating risk-scoring algorithms in the criminal justice system. Some of you will be familiar with this story already, so please forgive me for sharing it again. The story is this. For some years, an algorithm known as 'COMPAS' has been used in the US to rate how likely it is that someone who has been prosecuted for a criminal offence will commit another offence in the future. This rating can then be used to guide decisions regarding the release (on parole) of this person. The COMPAS algorithm is somewhat complex in how it works, but for present purposes, we can say it works like this: a risk score is assigned to a criminal defendant and this score is then used this to sort defendants into two predictive buckets: a 'high risk' reoffender bucket or a 'low risk' reoffender bucket.

A number of years back a group of data journalists based at ProPublica conducted an investigation into how this algorithm worked. They discovered something disturbing. They found that the COMPAS algorithm was more likely to give black defendants a false positive high risk score and more likely to give white defendants a false negative low risk score. The exact figures are given in the table below. Put another way, the COMPAS algorithm tended to rate black defendants as being higher risk than they actually were and white defendants as being lower risk than they actually were. This was all despite the fact that the algorithm did not explicitly use race as a criterion in its risk scores. This seems like a textbook case of indirect discrimination in action: we infer from the lack of fair representation in outcome classes that the algorithm must be relying on proxies that indirectly discriminate against members of the black population.

Needless to say, the makers of the COMPAS algorithm were not happy about this finding. They defended their algorithm, arguing that it was in fact fair and non-discriminatory because it was well calibrated. In other words, they argued that it was equally accurate in scoring defendants, irrespective of their race. If it said a black defendant was high risk, it was right about 60% of the time and if it said that a white defendant was high risk, it was right about 60% of the time. It turns out this is true. If you look at the figures in the table given you can see this for yourself. The reason why it doesn't immediately look like that is because there are a lot more black defendants than white defendants -- an unfortunate feature of the US criminal justice system that is not caused by the algorithm but is, rather, a feature the algorithm has to work around.

So what is going on here? Is the algorithm fair or not? Several groups of mathematicians analysed this case and showed that the main problem here is that the makers of COMPAS and the data journalists were working with different conceptions of fairness and that these conceptions were fundamentally incompatible. This is something that can be formally proved. The clearest articulation of this proof can be found in a paper by Jon Kleinberg, Sendhil Mullainathan and Manish Raghavan, which is like a version of Arrow's impossibility theorem for fairness.

The details are important and often glossed over. Kleinberg et al argued that there are three criteria you might want a fair decision procedure to satisfy: (i) you might want it to be well-calibrated (i.e. equally accurate in its scoring irrespective of social group); (ii) you might want it to achieve an equal representation for both groups in false positive ratings (more technically, you might want both groups to have the same average score in the positive class) and (iii) you might want it to achieve an equal representation for both groups in the false negative rating (more technically, you might want both groups to have the same average score in the negative class). They then proved that except in two unusual cases, it is impossible to satisfy all three criteria. The two unusual cases are when the algorithm is a 'perfect deterministic predictor' (i.e. it always get things right) or, alternatively, when the base rates for the relevant populations are the same (e.g. there are the same number of black prisoners as there are white prisoners). Since no algorithmic decision procedure is a perfect predictor, and since our world is full of base rate inequalities, this means that no plausible real-world use of a predictive algorithm is likely to be perfectly fair and non-discriminatory. What's more, this is generally true and not just true for cases involving risk scores for prisoners, even though this was the initial test case.

This result has significant practical implications for designers of decision-algorithms. It means they face some hard choices. They can have a system that is well-calibrated or one that achieves a fair representation, but not both. Plausibly, you might want different things in different decision contexts. For example, when deciding who would make a good doctor, you might want an algorithm that is well-calibrated: because you want some confidence that the people who end up becoming doctors are good at what they do and you want to stop people from assuming some people aren't good at what they do for irrelevant reasons. Contrariwise, when deciding who would make a good politician and should be put forward for election, you might want a system that achieves a balanced representation of the different social groups. This is to say nothing of the further complexities that arise from the fact that fairness is just one normative goal of social policy: there are other goals that can compete with it and crowd it out, e.g. security and well-being.

That's the first lesson to be drawn from the algorithmic fairness debate. What about the second? This lesson is that although there is a lot of concern about discrimination in decision algorithms, there is good reason to think that algorithmic decision procedures can be less discriminatory than traditional, human-led decision procedures. There are two reasons for this. The first is that we are prone to status quo bias when it comes to assessing the normative implications of any novel technology. We have a tendency to overemphasise the negative features of any new technology while neglecting the fact the current status quo is even worse. In this regard, I don't think it is controversial to say that human led decision-making systems are prone to bias and discrimination. This is a prevalent and systematic feature of them. This is in part because some people engage in direct discrimination, but also, more significantly, because many people engage in indirect discrimination of which they are completely unaware. We are prone to all manner of subconscious and automatic biases. We can work to counteract these biases, but only if we are aware of them and their effects on outcomes. We often aren't. This leads to the second reason for thinking that algorithmic decision-procedures might be less discriminatory than human-led ones. When designing algorithmic decision procedures we have to be explicit about the tradeoffs and compromises we are making, the datasets we are using, and the outcomes we are trying to achieve. This gives us greater awareness and control of its discriminatory properties. This is the case even though it is also true that certain features of the algorithmic decision process will be relatively opaque to humans [again, Kleinberg and his colleagues have a longish paper setting out a more technical argument in favour of this view - I recommend reading it].

3. Why We Might Want Robots to be Discriminatory
But what does all this mean for robotics? I want to close this talk by making two very brief arguments.

The first argument is that the lessons from the algorithmic fairness debate might mean a lot for robotics. The issues raised in the algorithmic fairness are particularly pertinent when an algorithm is used for general social decision-making. In other words, when the algorithm is expected to make decisions that might affect all members of the general population. This is the expectation for algorithms used to make decisions about credit risk, tax auditing and recidivism risk. To the extent that robots are designed to perform similar gatekeeping functions, they should be subject to the similar normative demands and will therefore face similar practical constraints (i.e. they will not be able to satisfy all possible fairness criteria at the same time). So, for example, imagine a security screening robot at an airport. That robot should be subject to the same demands of fairness and non-discrimination as a human screener (tempered by other policy aims such as security and well-being).

The second argument is that although this is true it might not be the normal case, particularly in the case of social robots. We could, of course, use social robots for general social gatekeeping, but I suspect the demand for this is going to be relatively limited. A lot of this admittedly hinges on how you define a 'social robot' of course, but I see social robots as embodied artificial agents, usually intended to participate in interpersonal social interactions. If you want a general social gatekeeper, then it's not clear why you would want (or need) to embody it in a social robot. This wouldn't be efficient or cost-effective. The embodied form is really only called for when you want a more meaningful, personal interaction between the human and the artefact. This might be the case for personal care robots or personal assistant robots. In those cases of meaningful, personalised interaction, the normative constraints of fairness and non-discrimination may not apply in the same way. This is not to suggest that we want robots to be racist or sexist -- far from it -- but we might not want them to be impartial either.

Think about it like this. Would you want your brother or best friend to be perfectly fair and non-discriminatory in how they interacted with you? No, you would want them to have some bias in your favour. You would expect them to abide by a duty of loyalty (or partiality) to your case. If they didn't, you would quickly question the value of your relationship with them and lose trust in them. If social robots are primarily intended to fulfil similar relationship functions (i.e. to provide companionship, care and so on) then we would probably expect the same from them.

This does not mean, incidentally, that I think robots should be used to fulfil similar relationship functions (i.e. be our friends and companions). I have my own views on this topic, but it is a longer debate, one that I don't have time for now. My only point here is that if they do perform such functions, it is plausible to argue that they should be bound by a duty of partiality or, to put it another way, a duty of positive discrimination toward their primary user.

So, in conclusion, although the algorithmic fairness debate does have some lessons for the robotic fairness debate, there may be important differences between them, particularly in the case of social robots.

Thank you for your attention.

Monday, May 20, 2019

#60 - Véliz on How to Improve Online Speech with Pseudonymity

Carissa Veliz

In this episode I talk to Carissa Véliz. Carissa is a Research Fellow at the Uehiro Centre for Practical Ethics and the Wellcome Centre for Ethics and Humanities at the University of Oxford. She works on digital ethics, practical ethics more generally, political philosophy, and public policy. She is also the Director of the research programme 'Data, Privacy, and the Individual' at the IE's Center for the Governance of Change'. We talk about the problems with online speech and how to use pseudonymity to address them.

You can download the episode here or listen below. You can also subscribe on Apple Podcasts, Stitcher, and a variety of other podcasting services (the RSS feed is here).

 Show Notes

  • 0:00 - Introduction
  • 1:25 - The problems with online speech
  • 4:55 - Anonymity vs Identifiability
  • 9:10 - The benefits of anonymous speech
  • 16:12 - The costs of anonymous speech - The online Ring of Gyges
  • 23:20 - How digital platforms mediate speech and make things worse
  • 28:00 - Is speech more trustworthy when the speaker is identifiable?
  • 30:50 - Solutions that don't work
  • 35:46 - How pseudonymity could address the problems with online speech
  • 41:15 - Three forms of pseudonymity and how they should be used
  • 44:00 - Do we need an organisation to manage online pseudonyms?
  • 49:00 - Thoughts on the Journal of Controversial Ideas
  • 54:00 - Will people use pseudonyms to deceive us?
  • 57:30 - How pseudonyms could address the issues with un-PC speech
  • 1:02:04 - Should we be optimistic or pessimistic about the future of online speech?

Relevant Links

Tuesday, May 14, 2019

Old Age and Decline: Some Philosophical Reflections

The Four Ages of Man - Nicolas Lancret

There’s an oft-repeated ‘fact’ thrown around in debates about retirement and old age. The details can vary but it’s something to the effect that when the pension entitlement age was set at 65 in the early part of the 20th century, very few people could expect to collect it, and those that did could only expect to collect for a few years (probably no more than 5). This was because life expectancy was so much lower back then. Hence setting pension entitlement at 65 was a relatively low cost gesture for the government. But what was low cost back then has turned into a major expenditure today, now that people are living so much longer and life expectancy has shot up. Whereas most people could only expect to live to their early 60s in the early 1900s, nowadays the majority can expect to live into their late 70s/early 80s. This places considerable strain on public finances and means more people are spending more of their lives in a ‘retired’ and ‘non-productive’ (from an economic/tax-paying perspective) state.

Having done some digging, it turns out this fact is not quite true. While it is true that life expectancy was much lower back then, that was mainly due to high infant and early adult mortality (due to infectious disease and war). If you cleared those early-life hurdles, and made it all the way to 65, you could expect to live a good bit longer, upwards of 13 years in fact (more if you were a woman). That post-65 life expectancy has gone up since then, but by much less than how much life expectancy as a whole has gone up. This doesn’t mean that costs are not increasing — the huge drop in early life mortality means a lot more people are making it to their late 60s. It also doesn’t detract from the fact that more and more people are entering this ‘retired’ phase of life.

But what does it mean to enter that phase of life? Many of us, myself included, have a negative perception of ageing and retirement. We see ‘old age’ as a period of inevitable decline and senescence. It is a phase of life marked by narrowing horizons, and a fall from grace and prowess. Although death looms large in old age, it is not the only negative aspect of the ageing process. If the Epicureans are right, then death itself is nothing to us; it is the period of life just before death — when we retreat from public view and lose our sense of significance, purpose and social meaning — that is the most existentially terrifying phase of life.

Is there anything to be said to quell these fears of ageing? Can we live valuable and meaningful lives in old age? There is a surprising lack of philosophical commentary on this issue. One of the more prominent contributors to the debate (Jan Baars) has argued that Western philosophy’s obsession with death has tended to suck attention away from old age. Nevertheless, there has been some work done on the topic and in what follows I want to share my own, poorly structured thoughts on it. This is a partial and selective take on old age, reflecting my own interests and biases. Still, some of you might find it interesting.

I start by looking at what I take to be the standard, ‘decline’-oriented view of old age. I then consider the alternatives to it.

1. The Decline Argument: A General Schema
There are some cultures where the elderly are afforded a lot of respect. Indeed, there is a famous adage stating that ‘with age comes wisdom’. Nevertheless, a common view in Western societies is that old age is a period of decline and devaluation. Simone de Beauvoir comments on this in her book The Coming of Age. She notes that many people are uncomfortable around the elderly. They see them as a social nuisance — a burden on the productive working population. They see them as something ’other’ or ‘foreign’. They try to marginalise and obscure them from sight through hollow gestures of charity:

Society appears to think that they belong to an entirely different species: for if all that is needed to feel that one has done one’s duty by them is to grant them a wretched pittance, then they have neither the same needs nor the same feelings as other men. 
(1972, p 9)

De Beauvoir links this attitude to the productivist ethos that underscores the modern economy:

The economy is founded upon profit; and in actual fact the entire civilization is ruled by profit. The human working stock is of interest only insofar as it is profitable. When it is no longer profitable it is tossed aside. 
(1972, p. 13)

This resonates. Certainly in debates about pension entitlements one often hears mention of the ‘burden’ that the elderly place on the working population. To be clear, this often comes with a sense of duty to the elderly, i.e. with a sense that they have done their bit for the economy and so deserve some protection, but, still, there is some begrudgery to the arrangement.

We should not, however, get too hung up on the economic devaluation of the elderly. It is significant but there is also a wider sense of decline and devaluation at play. There is the general belief that old age is a state in which you inevitably lose the capacities that make you valuable to yourself and your society (creativity, innovation, productivity, moral foresight, aesthetic beauty, physical prowess and so forth). Consequently, there is the sense that there is an inevitable general devaluation in old age.

This suggests that the following argument scheme undergirds the negative attitudes toward ageing:

  • (1) A life is valuable only if it has properties P1, P2…Pn. [The value premise]
  • (2) In old age, you inevitably lose (or experience some decline in) properties P1, P2…Pn. [The decline premise]
  • (3) Therefore, old age is inevitably a period in which your life becomes devalued.

I will evaluate the merits of this argument below. Before getting to that, however, it’s worth making a few comments on how it ought to be interpreted and understood.

First, note that this argument is a template that can be filled in with specific examples of the relevant value-conferring properties. You could, for instance, argue that life is valuable only to the extent that it is artistically creative; that old age inevitably brings about a decline in artistic creativity; and hence conclude old age results in an inevitable loss of value. Different combinations of value-conferring properties might make the argument more or less persuasive. Furthermore, these value-conferring properties could emanate from very different philosophical perspectives. For example, one could make the argument with personal value in mind (i.e. the value of a life to the one that is living it), or objective moral value in mind (i.e. the value of the life to the universe/humanity as a whole). Selecting one perspective over the other could make for very different arguments. A life could, after all, lack value from the personal perspective without lacking value from an objective perspective, and vice versa. This is to say nothing of socially constructed metrics of value (such as fame or economic value) and how they could be worked into the argument.

Second, note that there is some fuzziness to premise (2). This must be factored into the interpretation. I’ll say a bit more about the nature of ‘old age’ below, but here I want to point out that — at the limit — the decline premise is almost always true: people must lose some capacities in old age (after all, ultimately they must lose their lives). The only way this could fail to be true is if we invent perfect anti-ageing technologies that mean we can restore any lost capacity. This remains a pipe dream for now. This is important because, given the inevitable (at the limit) association between old age and loss of capacity, it might be tempting to simply define old age in terms of that loss of capacity. We must not succumb to that temptation. Doing so would make the argument trivially true. Old age must be defined as something other than the loss of properties P1, P2…Pn if the argument is to be interesting.

2. What is old age?
But then how should we define ‘old age’? We could lose a lot of time to this question. Jan Baars has some fascinating meditations on the ontology of old age in his work. He notes there are multiple different measures of old age and they don’t necessarily coincide on a common definition.

There is, for example, the standard chronometric measure of age. This is the measure of the number of minutes, hours, days, months and years since a person was born. This is a simple objective fact about the person that is easy to track and record. The problem is that this doesn’t tell you exactly when a person becomes old (if that even makes sense). You would have to pick some arbitrary cutoff point in the chronological measure (e.g. age 65 or 70) and define as ‘old age’ anything above that cutoff point. But that’s not particularly helpful since it doesn’t provide any reasoned justification for the choice of cutoff point.

This arbitrary, chronometric, approach to old age could then quickly lead to trouble. Here’s one: it is common for statisticians to associate clusters of capacities and abilities with chronometric ages (e.g. mental acuity, reading ability, physical dexterity). This allows them to say things like “If you are aged 18, you should expect to have properties X, Y and Z” and “if you are aged 65, you should expect properties P, Q and R”. But these are just statistical averages. You may not have those properties. This can lead to all sorts of odd statements being made about your age relative to the statistical average. For example, when I was younger it was common for students to be told their ‘reading age’ after standard assessments of reading comprehension. I remember I was quite chuffed when I heard that my ‘reading age’ was far in excess of my chronometric age. I was less chuffed, years later, when I was told that the age typically associated with my level of physical fitness was in excess of my chronometric age. This mismatch between one’s actual capacities and the statistical norms associated with chronometric age can lead to ageism, and makes articles like this one (on the ethics of ‘trans-ageism’) inevitable.

The other problem with the chronometric approach is that attitudes toward different chronometric ages are highly variable. The social and biological facts of ageing can change, depending on culture and technology. While 65 might have seemed like an appropriate retirement age 100 years ago, nowadays it doesn’t. That’s one reason why people call for increases in the retirement age (or a complete rejection of the concept). At the same time, as Baars notes, there is a tendency within certain groups to push back the chronometric cutoff to old age in order to promote certain interests. For example, an athlete is considered old in their 30s, ‘older workers’ are often relatively young (50ish), and so-called ‘mature students’ in universities can be very young indeed (early 20s).

As I say, we could waste a lot of time trying to figure out what old age actually is. I don’t want to go there because I don’t think there is a wholly satisfying answer. I do, however, think there are useful paradigm cases of old age that can guide our analysis. Thus, even though there is disagreement around the margins, I suspect most of us would agree that someone in their late 70s and 80s would count as being old age. Why so? I presume the answer lies in a combination of biological and chronometric reality and socially constructed norms and attitudes. Thus, I don’t think being old is a simple objective fact about a person — associated with their chronometric age — but rather is a complex bio-social-physical fact. It is not a fact that can be entirely self-determined (you cannot ‘will’ yourself to be younger), but it is a fact that is somewhat contingent and open to renegotiation (because of changing technological and medical realities as well as changing social perceptions).

With that clarification out of the way, I will spend the remainder of this article assessing the merits of the decline argument, focusing in particular on ways to object to its two premises. In doing so, I will have paradigmatic cases of old age in mind.

3. Rejecting the Decline Premise
One obvious way to object to the argument is to reject its second premise: the decline premise. The claim that old age is inevitably associated with some decline or obsolescence in value-conferring properties P1…Pn is, undoubtedly, going to be shaped by the statistical averages and social perceptions attached to certain chronometric ages. These averages and perceptions can be challenged.

To make this concrete, let’s consider a specific example. Suppose that within the world of mathematicians it is common to hear claims like “no mathematician over the age of 40 makes a significant breakthrough”. Any mathematician unlucky enough to be over the age of 40 (or should that be ‘lucky enough’ since the alternative fate is presumably worse?) would be devalued by other members of their profession as a result of this belief. But is the belief accurate? Any particular mathematician could undermine by pointing out that the belief does not hold true in her case (i.e. that they have made significant breakthroughs despite being over the age of 40), or by pointing out that it is based on a statistical misperception or error. In other words, they could either argue that (a) they are an exception to the perceived rule or (b) the perceived rule is false.

One or both of these strategies may work, depending on the individual case. In his article on successful ageing, Howard Harriott uses the example of the artist Matisse to illustrate how it is possible for an older person to live a life of significance. In Matisse’s case, his life’s mission centred around art and artistic creativity. He battled against the perception that art is a ‘young man’s game’ and dramatically illustrated how untrue this was through his own example. Despite being lambasted by the critics and suffering from several illnesses and frailties, he embarked on a ‘second life’ in his 70s and produced some of his most memorable work as a result:

In this late phase of his life, he embarks on a series of new works such as Florilége des Amours de Ronsard, Thèmes et variations, collages and innovative cutouts (papiers découpés). He embraces the “colors” of jazz as he transforms the vibrancy of jazz sounds and rhythms into a visual medium and produces his final triumph: the glorious chapel at Vence. Viewing Matisse’s later works, as for instance recently convened at the Musée de Luxembourg, in Paris, one gets the full sense of why Matisse’s work so illustrates the new paradigm of the creative life as seriously possible in old age. 
(Harriott 2006, 120)

Some people may object that Matisse and others like him are unusual figures — the exceptions that prove the general rule — but it’s not clear if that is accurate. It could be that general perceptions of decline in old age are misguided and that elderly people are much more capable than is believed. I have no doubt that there are a lot of ageist, unjustified assumptions made about them. Still, there are limits to this. For at least some value-conferring properties it will be true that old age is inevitably associated with decline and loss. For example, athletic prowess and physical fitness. At best, elderly people can minimise the losses they suffer with respect to those value-conferring properties: they cannot completely avoid them. So even if some formulations of premise (2) do not work; others probably will and the resultant decline and loss of value will be painful.

4. Finding alternative sources of value
A more promising strategy for objecting to the decline argument is to take issue with the first premise. Of course, as noted above, the first premise has no content in the abstract form. You need to identify specific value-conferring properties for it to make sense. Furthermore, it would be odd to reject all potential variations on the first premise: that would be tantamount to nihilism. If you want your life and the lives of others to have value you need to accept the existence of some value conferring properties. So what you need to do is find variations on premise (1) that are either more resilient to old age or not undermined or affected by old age.

This is what Howard Harriott recommends in his article on successful ageing. He argues that life has value to the one who lives it (personal value) when it is characterised by some commitment to ideals. These ideals can take many forms, e.g. commitment to artistic excellence, scientific discovery and so on. If you want to retain personal value into old age then you need to focus on ideals that can sustain your commitment into old age. Again, Matisse is Harriott’s go-to example, but others easily spring to mind. I think a lot of Einstein in this regard. He remained steadfastly committed to developing a unified theory of physics right up until his final days. According to reports, he was scribbling equations in his hospital bed just hours before he died. Admittedly, Einstein’s unified theories weren’t successful in the objective sense, but at least they gave him a sense of purpose right up until his death. He was committed to an ideal — scientific discovery — that was relatively impervious to the vicissitudes of old age. It is also worth mentioning here that studies that have been done on elderly populations suggest that they derive most meaning from sustained social and family relationships, both of which can be sustained despite ageing (although both can suffer too).

The important lesson to learn from both Einstein and Matisse is that retaining value in old age is a function of choosing ideals that are resilient. But what if your ideals are not so resilient? What if, for example, your ideals are built around physical fitness and athletic prowess? The answer is that you could probably sustain some version of those ideals into old age, but it might require some modification. If you were once an elite athlete, you will have to accept that you won’t be able to compete to the highest levels into your dotage. But you could continue to be the best within your age range (injuries permitting) or you could switch to training and educating future generations of athletes. In other words, if you have some flexibility of mind, you can sustain value in the face of changing circumstances.

This last point is worth emphasising. In classic Stoic and Epicurean philosophy, people were encouraged to ‘thanatise’ their desires so as not to be so afraid of death. In other words, they were encouraged to accept the unavoidability of death and factor that into how they structured and planned their lives. Jan Baars recommends a similar strategy when it comes to ageing. He thinks we need to accept that everything in life (peak productivity; cognitive capacity; physical prowess etc) is finite and subject to decay. We need to build a conception of a meaningful life that recognises and makes space for that finitude.

That sounds good in theory, but might be hard in practice. One reason it might be hard is because this whole line of argument assumes that people can simply pick and choose their own values rather than having them imposed from the outside. The great tragedy of ageing in the modern world is that devaluation results from the imposition of values and standards from the outside. How do you deal with that problem?

5. The Value of Escaping Imposed Ideals
One way to cope with that problem is to take solace in the fact that if you are no longer perceived to be valuable in the eyes of others you are both (a) more free to determine your own value in life and (b) exempted from the burdens and expectations that are imposed on younger people. This can be liberating and uplifting.

Consider once more the productivist ethos that pervades much of modern life. According to this ethos, you are valuable only to the extent that you make some productive contribution to society. This could take a number of different forms, but for most people it is economic productivity that matters most. While making economic contributions can be very meaningful to some people, it can also be exhausting and dispiriting. Instead of following your passions and talents, you have to fit within the demands of the labour market. You have to do something ‘useful’ and avoid idle luxuries. You have to compromise on your values and sell yourself to others. You have to impress them and suck up to them. You have to ingratiate yourself with the powerful and shower them with false praise. In return, they might do the same for you. As a result, you all benefit from increased perception of social value. But at what personal cost? No longer being seen as productively valuable might give you a nice excuse to be rid of all this fakery and flattery.

There could also be a significant gendered aspect to this. Simone de Beauvoir comments on the different expectations of age in her work. And while researching this article, I randomly came across a piece written by the philosopher Andreas Blank about ageing and self-esteem in the writings of Anne-Thérèse de Lambert. Lambert was an 18th century French intellectual and essay writer who wrote about the ‘economy of self-esteem’ and ageing in women. Blank argues that her views provide a contrast with those of the well-known maxim-writer La Rouchefoucauld. Whereas he essentially accepted the modern view that old age was a period of decline from former prowess, she argued that it could be liberating, particularly for women. In a society that did not value women for intellectual ability or economic productivity, but essentially valued them only for looks, charm and fertility, there was something to look forward to in old age. Freed from the burden of erotic expectation, and from the need to impress powerful men, women could cultivate a more intellectual and satisfying mode of life. As she put it:

[Old age] liberates us from the tyranny of opinion. When one is young, one only dreams of living in the idea of someone else; one must establish one’s reputation and create for oneself an honorable place in the imagination of others, and be happy even in their idea; our happiness is not at all real, it is not ourselves whom we consult but others. In a different age, we turn to ourselves, and this return has sweetness, we begin to consult ourselves, and to believe ourselves; we escape chance and illusion; men have lost their right to deceive us… 
(Lambert, quoted in Blank 2018, 299-300)

This is, no doubt, naive and optimistic but there is something very appealing in what she has to say. The idea that old age can free us from the ‘tyranny of opinion’ is one that I find comforting.

Monday, May 13, 2019

AI and Sexuality (New Paper)

I have new paper. This one is set to appear in the Oxford Handbook of the Ethics of AI, which is edited by Markus Dubber, Frank Pasquale and Sunit Das. The book isn't out yet. I believe it is due out in the Autumn/Fall. You can access the penultimate draft at the links below.
Title: Sexuality
Book: The Oxford Handbook of the Ethics of Artificial Intelligence
Links: Philpapers; Researchgate; Academia
Abstract:  Sex is an important part of human life. It is a source of pleasure and intimacy, and is integral to many people's self-identity. This chapter examines the opportunities and challenges posed by the use of AI in how humans express and enact their sexualities. It does so by focusing on three main issues. First, it considers the idea of digisexuality, which according to McArthur and Twist (2017) is the label that should be applied to those 'whose primary sexual identity comes through the use of technology', particularly through the use of robotics and AI. While agreeing that this phenomenon is worthy of greater scrutiny, the chapter questions whether it is necessary or socially desirable to see this as a new form of sexual identity. Second, it looks at the role that AI can play in facilitating human-to-human sexual contact, focusing in particular on the use of self-tracking and predictive analytics in optimising sexual and intimate behaviour. There are already a number of apps and services that promise to use AI to do this, but they pose a range of ethical risks that need to be addressed at both an individual and societal level. Finally, it considers the idea that a sophisticated form of AI could be an object of love. Can we be truly intimate with something that has been 'programmed' to love us? Contrary to the widely-held view, this chapter argues that this is indeed possible.

Thursday, May 9, 2019

#59 - Torres on Existential Risk, Omnicidal Agents and Superintelligence

Phil Torres

In this episode I talk to Phil Torres. Phil is an author and researcher who primarily focuses on existential risk. He is currently a visiting researcher at the Centre for the Study of Existential Risk at Cambridge University. He has published widely on emerging technologies, terrorism, and existential risks, with articles appearing in the Bulletin of the Atomic Scientists, Futures, Erkenntnis, Metaphilosophy, Foresight, Journal of Future Studies, and the Journal of Evolution and Technology. He is the author of several books, including most recently Morality, Foresight, and Human Flourishing: An Introduction to Existential Risks. We talk about the problem of apocalyptic terrorists, the proliferation dual-use technology and the governance problem that arises as a result. This is both a fascinating and potentially terrifying discussion.

You can download the episode here or listen below. You can also subscribe on Apple Podcasts, Stitcher and a variety of other podcasting services (the RSS feed is here).

Show Notes

  • 0:00 – Introduction
  • 3:14 – What is existential risk? Why should we care?
  • 8:34 – The four types of agential risk/omnicidal terrorists
  • 17:51 – Are there really omnicidal terror agents?
  • 20:45 – How dual-use technology give apocalyptic terror agents the means to their desired ends
  • 27:54 – How technological civilisation is uniquely vulernable to omnicidal agents
  • 32:00 – Why not just stop creating dangerous technologies?
  • 36:47 – Making the case for mass surveillance
  • 41:08 – Why mass surveillance must be asymmetrical
  • 45:02 – Mass surveillance, the problem of false positives and dystopian governance
  • 56:25 – Making the case for benevolent superintelligent governance
  • 1:02:51 – Why advocate for something so fantastical?
  • 1:06:42 – Is an anti-tech solution any more fantastical than a benevolent AI solution?
  • 1:10:20 – Does it all just come down to values: are you a techno-optimist or a techno-pessimist?

Relevant Links