'Fairness and Abstraction in Sociotechnical Systems' (ACM Conference on Fairness, Accountability, and Transparency (FAT*) by Andrew D. Selbst, danah boyd,
Sorelle Friedler, Suresh Venkatasubramanian
and Janet Vertesi
comments
A primary goal of the FAT* community is to develop machine-learning based systems that, once introduced into a social context, can produce social and legal goals such as fairness, justice, and due process. Bedrock concepts in computer science such as abstraction and modular design are used to define notions of fairness and discrimination, to produce fairness-aware learning algorithms, and to intervene at different stages of a decision-making pipeline to produce "fair" outcomes. In this paper, however, we contend that these concepts render technical interventions ineffective, inaccurate, and sometimes dangerously misguided when they enter the societal context that surrounds decision-making systems. We outline this mismatch with five "traps" that fair-ML work can fall into even as it attempts to be more context-aware in comparison with traditional data science. We draw on studies of sociotechnical systems in Science and Technology Studies to explain why such traps occur and how to avoid them. Finally, we suggest ways in which technical designers can mitigate the traps through a refocusing of design in terms of process rather than solutions, and by drawing abstraction boundaries to include social actors rather than purely technical ones.
'Fat horses and starving sparrows' by Rebecca Giblin in (2018) 232
Overland comments
‘One of the most salient features of our culture is that there is so much bullshit’ – so begins moral philosopher Harry Frankfurt’s treatise on bullshit and its function. Bullshit comes, he argues, from one who ‘does not care whether the things he says describe reality correctly’, but says them regardless, in pursuit of their desired ends.
Bullshit has been enjoying unprecedented success in recent years. The Mexicans will pay for the wall. Britain’s National Health Service will enjoy a weekly £350m injection after Brexit. Australia stops the boats to prevent drownings at sea. But the bullshit I’m interested in right now is that populating Australia’s copyright reform debates.
A great deal of this bullshit is motivated by good intentions – most notably, the desire to sustain writers’ incomes in an era of precipitous, disastrous decline. In the last major survey, conducted by Macquarie University researchers in 2015, Australian authors were found to earn an annual average of just $12,900 from their writing work; the median, at $2,800, is even more concerning. In the UK, which has better longitudinal data, earnings of professional writers have dropped 42 per cent in real terms between 2005 and 2017, according to the Authors’ Licensing and Collecting Society. In that same time, the proportion able to make a living solely from writing work fell from 40 per cent to 13.7 per cent. The jobs at newspapers and magazines that used to so often be relied on to bolster book earnings have largely evaporated. Many of us know people who have lost their writing jobs, or who are just barely clinging on. Much of the blame is aimed at Google and Facebook, and understandably so: the two companies have managed to vacuum up some 60 per cent of global online ad revenue, despite investing almost nothing in producing the content to which it’s attached.
But pure intentions don’t stop bullshit from being bullshit. Frankfurt’s theory doesn’t require nefarious aims – simply the saying of things, regardless of truth, in pursuit of desired ends. As he puts it, ‘[t]he bullshitter … does not reject the authority of the truth, as the liar does, and oppose himself to it. He pays no attention to it at all.’
Frankfurt sees bullshit as the greatest enemy of truth. I don’t know if I agree, but I do think bullshit is dangerous. In this case, the bullshitters’ indifference to facts is hurting the interests of the authors they are ostensibly trying to protect.
So what is the bullshit I’m talking about? And what are the issues and interests it serves to obscure?
Let’s start with ‘free is not fair’, the catchy slogan of a campaign led by the Copyright Agency, which has been used to argue against the proposed adoption of a ‘fair use’ exception to copyright. For those who haven’t been following, there has been fierce debate over whether Australia ought to adopt a flexible exception to copyright infringement that would permit, without payment, any kind of use, so long as it’s ‘fair’. This is already the law in countries such as the US, with ‘fairness’ determined by taking into account all relevant factors, including the effect of the use on the potential market for or value of the copyright material.
Many people claiming that ‘free is not fair’ know a lot about copyright. They know that our law has permitted, for over a century, fair use for purposes such as research, criticism/review and news reporting. And, unless you think you should pay for every use of every copyrighted thing – for example, every time you access a webpage, use a search engine, forward an email or retweet a photo – you know free can be fair. The slogan even contradicts itself, since under a fair-use exception, if the use of a work for free is not fair, it wouldn’t be permitted. That is the entire point. Thus ‘free is not fair’ is textbook bullshit.
The widespread claim that the Productivity Commission recommended copyright be reduced from its current period (life of the creator plus seventy years) to a flat twenty-five years is bullshit, too. I’ll come back to what the commission did say shortly, but for now it’s sufficient to say that it made no such recommendation. In fact, the commission explicitly stated that international law would prevent it from making such a change. The publishers who pushed the claim knew all this, but ran with it anyway.
Of course, there is also bullshit coming from the other side of the debate. Those seeking to expand user rights sometimes make blithe assurances that this will have negligible impact on uses that are currently paid. Fair use certainly should work that way, since courts would be required to take into account potential market harms in determining whether a use is ‘fair’ at all. But the distinct lack of emphasis on safeguards or guarantees can justifiably give rise to concerns that those making the claims do not ‘care whether the things he says describe reality correctly’ – especially if they come from those who are already gobbling up your lunch.
Counterintuitively though, it’s the ostensibly pro-author claims that risk harming authors the most. Those arguing to maintain the current regulations may see these slogans as benign simplifications. After all, copyright is arcane and complex. Free is not fair is a simpler sell than Free is sometimes fair, but we worry judges might find things to be fair when we think they should be paid for.
Similarly, The Productivity Commission wants to take away your copyrights is a proxy for We think its recommendations are against our interests in ways that are too complex to explain, so we are going to arouse your outrage against a non-existent threat to undermine what else it had to say.
However well intentioned these claims might be, they sell authors short. The reasons why can be found in the two fundamental rationales for granting copyright in the first place. ...
Giblin goes on to comment
Australia’s current approach to protecting authors is a manifestation of trickledown economics, that theory of horses and sparrows: feed the horses enough oats and some will fall through to feed the birds. There are plenty of oats. By 2016, according to their own publicly released data, publishing behemoths Penguin Random House and Simon and; Schuster had seen their profit margins grow to 16 per cent. Yet a recent study, The Contribution of the Publishing Industry to the UK Economy, estimated that just 3 per cent of earnings went to the authors from whose minds sprang those rivers of gold. In other words, we have fat horses and starving sparrows.
This is why bullshit in the copyright debate is dangerous. By conflating authors’ interests with those of investors, we obscure the important differences between them. And it distracts us from the questions we should be asking – such as how can we secure authors a fairer share, while also maintaining incentives.
