The implicit premise here is that one can effect change via ethical training. In some respects, this is possible. For example, professional training e.g. against corruption might well have some success. We teach doctors and social workers about safeguarding too.
But many other problems are not amenable to such treatment. For example, anti-corruption training is unlikely to work in an environment where there are few penalties against corruption. The average British civil servant is generally convinced not only that bribery is bad but also that one who acquiesces to it may well be caught and humiliated. The Indian Administrative Service officer who is told not to accept bribes will probably laugh off such warnings. Doctors’ professional training doesn’t seem to make them immune to being lobbied by drug companies.
The implicit premise here is that the lacuna in ethical training here falls into the former bucket. I do not think that this is so. The major difference between the two categories is a question of compatibility with the desires of those in power. Most governments don’t want their civil servants to be corrupt, and, given the resources, will work with some success towards that goal. By contrast, large technology companies’ business models rely on violating privacy—it’s not just a procedural failing that can be eradicated or successfully fought with training.
When the incentives aren’t right for the powerful, data scientists, software engineers &al. will largely follow the money. At best, such ethical training programmes might make it a bit more expensive to hire people for unethical ends. But when one thinks of the major crimes, say, of ML, it seems unlikely that expense alone is going to be a problem. Would the Chinese government stop using ML to discriminate against Uyghurs by detecting ethnicity using facial recognition simply because they have to pay more? Would Google give up on billions in advertising revenue because they’d have to pay ML wallahs a few hundred thousand more a year? For that matter, would their shareholders want to fire all these people to save far less than they bring in in profit?
The largest counterexample I can think of is in the field of law. Cynics are right to a certain extent: many lawyers are pretty mercenary about their jobs. But there is a kind of ethical core to it; one thinks of, say, the Hong Kong Bar Association, who under tremendous pressure continue to try to speak truth to power, or human rights lawyers in Burma and China who at grave risk to their lives, livelihoods and families insist upon trying to see that justice is done. The difference here is that they can try to make a difference alone. A human rights lawyer might win one case. A liberal judge in apartheid South Africa can stop evictions of blacks simply on the basis of their skin colour under the Group Areas Act. But as above, the best case for data science is that some withhold their labour and salaries go up. Perhaps some will be motivated to build tools to resist unethical practices too, in the same way that e.g. developers can contribute to Signal. However, I suspect that most who would be inclined to do this would do so anyway.
So far this has merely been an argument against viewing ethical training as a panacea. But I think it’s more dangerous than that. Effort spent on trying to ensure that there’s ethical training in these courses could be spent lobbying politicians or campaigning amongst the public. That can only really be caused by a change in popular ideology and government policy; sometimes, the mores of the powerful will restrain their excesses, and when that’s inadequate, governments can help. As a mechanism of effecting social change generally, ethical training for professionals is severely hobbled. I won’t say that it’s useless, because it’s not, but I do think that choosing this over a direct struggle over public opinion and government policy is misguided.
But many other problems are not amenable to such treatment. For example, anti-corruption training is unlikely to work in an environment where there are few penalties against corruption. The average British civil servant is generally convinced not only that bribery is bad but also that one who acquiesces to it may well be caught and humiliated. The Indian Administrative Service officer who is told not to accept bribes will probably laugh off such warnings. Doctors’ professional training doesn’t seem to make them immune to being lobbied by drug companies.
The implicit premise here is that the lacuna in ethical training here falls into the former bucket. I do not think that this is so. The major difference between the two categories is a question of compatibility with the desires of those in power. Most governments don’t want their civil servants to be corrupt, and, given the resources, will work with some success towards that goal. By contrast, large technology companies’ business models rely on violating privacy—it’s not just a procedural failing that can be eradicated or successfully fought with training.
When the incentives aren’t right for the powerful, data scientists, software engineers &al. will largely follow the money. At best, such ethical training programmes might make it a bit more expensive to hire people for unethical ends. But when one thinks of the major crimes, say, of ML, it seems unlikely that expense alone is going to be a problem. Would the Chinese government stop using ML to discriminate against Uyghurs by detecting ethnicity using facial recognition simply because they have to pay more? Would Google give up on billions in advertising revenue because they’d have to pay ML wallahs a few hundred thousand more a year? For that matter, would their shareholders want to fire all these people to save far less than they bring in in profit?
The largest counterexample I can think of is in the field of law. Cynics are right to a certain extent: many lawyers are pretty mercenary about their jobs. But there is a kind of ethical core to it; one thinks of, say, the Hong Kong Bar Association, who under tremendous pressure continue to try to speak truth to power, or human rights lawyers in Burma and China who at grave risk to their lives, livelihoods and families insist upon trying to see that justice is done. The difference here is that they can try to make a difference alone. A human rights lawyer might win one case. A liberal judge in apartheid South Africa can stop evictions of blacks simply on the basis of their skin colour under the Group Areas Act. But as above, the best case for data science is that some withhold their labour and salaries go up. Perhaps some will be motivated to build tools to resist unethical practices too, in the same way that e.g. developers can contribute to Signal. However, I suspect that most who would be inclined to do this would do so anyway.
So far this has merely been an argument against viewing ethical training as a panacea. But I think it’s more dangerous than that. Effort spent on trying to ensure that there’s ethical training in these courses could be spent lobbying politicians or campaigning amongst the public. That can only really be caused by a change in popular ideology and government policy; sometimes, the mores of the powerful will restrain their excesses, and when that’s inadequate, governments can help. As a mechanism of effecting social change generally, ethical training for professionals is severely hobbled. I won’t say that it’s useless, because it’s not, but I do think that choosing this over a direct struggle over public opinion and government policy is misguided.