‘Invisible Women – Exposing Data Bias in a World Designed for Men’ by Caroline Criado Perez

Review by Arran Pearson

In 2008, Chris Anderson the (then) Tech Editor for Wired Magazine wrote an article titled ‘The End of Theory: The Data Deluge Makes the Scientific Model Obsolete’.  The main thrust of the article was that with the ubiquity of data, models and theories are essentially obsolete as it was possible to get the answers to pretty much everything through simple correlation.

This article was hugely influential and shaped a lot of the current thinking about data science and modelling. Caroline Perez’s book serves as an effective counterpoint to the article underscoring (with hard facts and figures) how belief in ‘the data’ not only results in a range of sub-par outcomes for women but also highlights a general issue with the way unconscious bias can creep into even the most objective of models.

Caroline Perez’s premise is that when we say ‘person’ the default is that person = man.  This assumption is then acted on in ways that result in poor outcomes for women specifically and society generally.  As an example, bulletproof vests are designed to be ‘gender neutral’ however the default body shape is male and for a couple of obvious reasons the neutral fit either fits women poorly and doesn’t provide the required protection or is not worn at all.  Both result in more women being injured which also feeds into the stereotype that women don’t belong in ‘front-line’ policing / combat roles due to the higher prevalence of injury.

What I found really fascinating was that Caroline Perez has dug beneath simple examples such as emergency housing designed by a committee of men that did not include any place to cook or prepare food (it never occurred to them that this was important).  Rather she has looked at the reasons why the data bias might exist and tried highlight some of the specific issues that need to be addressed or even acknowledged that the bias exists.

One of the important takeaways for me is that in the increasingly data-driven society, much use is made of large underlying ‘reference sets’ which then guide the way in which the machine will respond.  For example, speech recognition works because there is a reference set containing thousands of recordings of how words are said. Unfortunately for women, even then most ‘neutral’ of these reference sets are over 70% male which comically results in poor outcomes by Siri and (potentially) tragically as speech recognition gains more of a foothold as a tool for emergency and essential services.

I would say that most men of my generation have been raised and surrounded by amazingly capable women in both our professional and personal lives and it is sometimes difficult to understand why gender disparity (in pay, promotion, share of unpaid work – the list continues) exists.  This book offers a reasoned explanation for some of the biases that underlie why Women often fail to achieve their potential despite (from a male perspective) a society that empowers them to succeed.  Indeed several examples in the book highlight the ways that men have tried (and failed) to improve outcomes for women in a way that, ironically, never involved seeking women’s input.

At times I found this quite a challenging read as it forced me to confront a range of my own assumptions and understanding (from the male perspective the sections on unpaid work will make you uncomfortable).  But then, challenging the status quo is rarely pleasant.  

I’d recommend this book as a must read for anyone generally but to Men in particular.  I suspect many of the examples will have women readers shouting ‘Duh!’ but I found several examples of things that I ‘knew’ but just hadn’t spent the time working things through to logical conclusions to understand why.  A book for everyone’s ‘must read’ list.