Fembots, sexbots, killerbots, chatbots: all these and more are the subject of Tracey Spicer’s latest book Man-Made: How the bias of the past is being built into the future. But while the book offers an overview of the problem just as AI dominates the news, its critical analysis falls short.

Spicer has collected a vast range of research and case studies, and interviewed some genuinely impressive people in order to map the many ways in which bias infiltrates the technology we use—or is used upon us—everyday. Man-Made’s premise is that societal biases, particularly sexism, are being embedded into technology, with a particular focus on artificial intelligence. This often happens as a result of the systems learning from biased training data, which reflects and amplifies real-world current and historical inequalities. Within each chapter Spicer takes on an area of enquiry, painting cheeky vignettes that have been praised for being ‘accessible to all people’ in a ‘fabulously conversational style.’ She quips ‘Preach, sister!’ when referencing Caroline Criado-Perez’s Invisible Women on how the world is designed for men; ‘Golly, gee, it’s our fault yet again’ in response to being told to speak louder when speech recognition systems don’t work for women; and ‘Crikey: technical language alert!’ when she, a self-described ‘digital newbie’, gets into the weeds.

There is no corner of the sandpit that Spicer is unwilling to dig in. While it is clear that Man-Made takes seriously the issues she presents, however, what is decidedly un-serious are the solutions and conclusions drawn. Man-Made lacks a meaningful framework for understanding the injustices exacerbated by the tech industry beyond a shortage of diversity in male-dominated fields. Spicer has an evident knack and passion for collating the abundant examples of biased tech, but despite having read widely on the topic, she keeps coming back to the same solution: just add women. Credulously, she has fallen into a liberal feminist trap of believing that improving diversity will produce the structural change required to make a meaningful difference to our digital future.

The lack of diversity in the tech industry is undoubtedly a problem, and the task of increasing representation of women is generally presented in two ways. Firstly, as a way for women to climb the economic ladder, to have the same financial and professional opportunities as men. Secondly, as the means to improve the technology sector and its outputs. The theory goes that, by increasing diversity—of employees and therefore presumably perspectives, experiences and ideas—we will achieve better, more ethical tech.

The trouble is that, in order to design and develop ethical, socially beneficial technology, we need to address the imperatives of the industry (wealth accumulation) and the ideological structures upon which it has been built (patriarchal white supremacy). There is nothing to guarantee that improving diversity will produce this kind of structural change.

From the perspective of large companies and firms, however, diversity is a golden ticket. Corporate calls for Diversity, Equity and Inclusion strategies most commonly argue in its favour not because of some commitment to fairness, but because it’s good for business. Research conducted by consulting firms tells us that diversity matters, because diversity wins. Winning here means, of course, increased productivity, output and financial performance.

Feminist technoscience researcher Dr Thao Phan emphasised in a Liminal interview that she does not believe that diversifying the tech sector will lead to better implementation of technologies. ‘The aim of diversity and inclusion strategies isn’t to challenge the dominance of technocratic forms of governance,’ she says, ‘[it] is to further entrench and legitimise that regime.’ Similarly, in On Being Included, Sara Ahmed offers an essential critique of institutional diversity tactics as a method of not doing anything, by appearing to do something. In a 2015 speech, feminist political activist Angela Davis said that she has a hard time accepting diversity as a synonym for justice, seeing it as a corporate strategy designed to ensure the institution functions in the same way it always has.

Diversifying the police, for example, does not make it less violent or racist, because the police is structurally violent and racist. Diversifying the tech industry as it currently stands faces a similar challenge, because the industry is fundamentally built upon an ideology of ruthless capitalistic gain, perpetual growth and value extraction. Where Man-Made fails is in its commitment to the fantasy of forcing systemic change by fixing the tech talent pipeline. Efforts to increase diverse representation in the tech industry can have value, but we should be realistic about their ability to lead to fairer, more ethical technical outcomes. We should heed Ahmed’s advice: which is not that we should stop pursuing diversity, but that ‘we need to keep asking what we are doing with diversity.’

If diversity isn’t enough to solve tech’s problems, then what is? Man-Made comes tantalisingly close to important underlying issues, but the analysis falls short. When discussing algorithmic injustice, for example, readers are presented with COMPAS—a thoroughly racist algorithmic system used in the US to predict what people convicted of a crime are most likely to reoffend, therefore recommend longer sentences. This is followed by an examination of gendered and racial biases in facial recognition technology, and how it results in people being wrongly arrested and detained, or used to uphold oppressive regimes. Spicer goes so far as to rebuke the deeply transphobic area of gender recognition, in which biometric models are used to attempt to predict a person’s gender—a horrible example of pseudoscience commonly compared to technological phrenology.

The impact of these systems can mean life or death, and requires urgent attention. And yet, instead of encouraging readers to consider the need to dismantle racialised surveillance, techno-carceral ideologies, and the system of mass data generation and commodification enabling all of it, we are told to take responsibility for where we find ourselves.

How often do you blithely share a selfie? … If we’re going to accuse the villains, we might as well sheet some of the blame ourselves.

To suggest that individuals are responsible for punitive facial surveillance and predictive-policing because they wanted to share photos with their friends, only for those images to be scraped without knowledge or consent into facial databases, plays directly into the hands of the firms who stand to benefit from this tech the most.

Spicer goes on:

There’s little we can do about breaches of privacy, aside from being wary of sharing photos online. But there is something we can do about the diversity of the images used to train our robot masters.

Leaving aside the unnecessarily defeatist comment on privacy, what Spicer alludes to here is that by improving the variety of faces in the datasets, the models will improve their accuracy—an assumption which has been questioned by many, including academic Kate Crawford in Atlas of AI. But even if facial recognition were perfect, it would still be ready to be employed as punitive surveillance by police and corporations. Using the example of Dr Joy Buolamwini’s pivotal research into racial bias in this kind of software, Spicer exclaims: ‘The implications for people of colour are shocking: The technology that will be ruling the future doesn’t deem you to be human.’ The question as to whether this technology should rule our future is taken as a given. The focus on bias, while an important facet, only addresses part of the issue.

Spicer also circles around the military industrial complex and flirts with its relationship to techno-capitalism. The book articulates well that

ethics are being sacrificed as world powers endeavour to keep pace with one another. Microsoft and Amazon work closely with the Pentagon. What’s their primary motive? Profit.

This is the crux of it, yet it would hold more power without the further assertion that ‘the way forward is to improve diversity in military decision-making and peacekeeping.’

Still, the scope of issues covered is commendable and Spicer makes it clear that bias in AI systems is not a distant threat, but is already causing harm. After Man-Made makes a convincing case that a problem exists, readers are primed to be presented with solutions. At this point, however, the playful tone that served to draw them into the earlier parts of the book devolves to a somewhat panicked frenzy, as the author reveals her apparent bewilderment about what should happen next by throwing ideas in rapid fire at the wall, to see what sticks.

Introducing a universal basic income and increasing taxing large tech companies more are two such ideas, and would hold more weight if we hadn’t also been assured by Spicer that she’s not peddling some ‘lefty commie utopia.’ Another ‘easy suggestion’ is to include a human in the loop in AI systems. This may indeed seem easy—a harmless comment that sounds good, progressive, and uncontroversial. What it misses however, is how the concept of a human in the loop is neither easy, nor uncontroversial, nor necessarily even effective as an AI governance model.

Spicer further suggests that it’s ‘time to bankroll tech companies run by women and people from marginalised populations.’ Examples include the Australian company Canva, or Spicer’s ‘latest mission’ to switch from Uber to women-run rideshare app, Sheebah. It’s true that venture capitalist funding does leave women-led technology initiatives behind. In 2022, startups founded by women raised only 1.9 per cent in venture capital in the US. But instead of asking how we can get more women-led businesses recognised by venture capitalists, we should be questioning whether this model, and the mythology of innovation that surrounds it, is actually the driver of progress best suited to socially responsible and useful technology.

As emphasised by Dr Jathan Sadowski, venture capital is an engine not of innovation, but of capitalist innovation. Socially progressive innovation often remains unfunded because it is unprofitable. Tinkering around the edges of such a system—say, by encouraging investment in more women-led businesses—will not solve the problem of capital doing what capital does best.

According to the promotion by publisher Simon & Shuster, ‘Walkley Award-winning journalist Tracey Spicer exposes the next frontier of feminism.’ So we should ask, what kind of frontier is being ‘exposed’ here? Because people have been ringing the alarm on these issues for close to a decade. In the epilogue, the author informs us: ‘this is frontier feminism. After reading this book, your consciousness is raised.’ If she says so! But what exactly does this ‘frontier feminism’ mean? Spicer clearly recognises that intersectionality is a thing, and regularly seeks to expand the scope of her enquiry beyond just women to also include people of colour, Indigenous groups, and people with disabilities. Yet this does little to assuage the dominant throughline of liberal feminism seeking to work within the status quo to integrate women into the existing structures.

In her book by the same name, Koa Beck defines white feminism as an approach toward achieving gender equality that focuses on individual accumulation, capital and individuality—accruing power without reconsideration or redistribution. In this conception, white feminism sits in the same family as girlboss feminism or corporate feminism. As an ideology, it doesn’t demand that institutions or powerful bodies are forced to change. Rather, the strategies put forward focus on enabling women to excel within the system as it currently exists. A classic—and controversial—example is former Facebook executive Sheryl Sandberg’s 2013 book Lean In, famously criticised by bell hooks as an example of ‘faux feminism.’

What we need instead is to reimagine power, how we understand it, and how we exercise it. In patriarchy and white supremacy, power is understood as power over someone else, accumulated through domination. (Notably, Spicer’s promotional call to action is that ‘we need to master this technology before it enslaves us.’) The idea of domination and exploitation is fundamentally intertwined with big tech and the purveyors of the codified bias called into question in Man-Made. In order to make meaningful change, we should heed instead a lesson from Mona Eltahawy in Seven Deadly Sins for Women and Girls: ‘we must dismantle the hierarchies that patriarchy uses, not aim to climb our way up its ladder of injustices.’ While Spicer does acknowledge that ‘obviously, the patriarchy is a structure,’ it is referenced to assure readers that ‘this is not a rant against individual men,’ or later, to assert that ‘of course, artificial intelligence is not, of itself, evil; nor are the men who dominate the industry.’ Where Man-Made falls short is that it does not ask readers to critically consider the structure and ideologies of the tech sector, but instead supposes that this framework should be ascended and made equal.

Despite its shortcomings, Man-Made does promise to bring the long-established issue of bias in tech to a broader audience. No social movement has ever been successful by staying small, niche or pure of politics. As political organiser Yotam Marom has highlighted, tensions rise when a cause starts to gain momentum, but movements ultimately suffer if they don’t understand what it takes to build popular power. So, if Man-Made provides some readers with a useful entry point to consider the issues of bias in tech and the need for change: hello, you are welcome in this movement, we need you here. But also: please, do not stop here. If we truly want to create a liberatory digital future, we need to think much more radically.

Image: a detail from the cover of Man-Made