WHAT’S IN THE BOOKS: WHAT COUNTS WHEN YOU’RE COUNTING?
Sometimes what you take away from a book is not at all what it set up to do in the first place. You are caught off guard by a sentence, a clause, a paragraph and while your eyes skim over the rest of the pages your mind keeps lingering with that one idea that struck you.
That happened to me with Deborah Stone’s Counting: How We Use Numbers to Decide What Matters. Stone, an American political scientist, who is mainly known for her award-winning book Policy Paradox wants us to have a second look at statistics and data and what they do to the world. Most of it will be familiar terrain for everyone working in data, and she won’t really shake you to the core with her thesis that numbers are not as objective as they pretend to be.
Counting is losing
In six chapters — and an epilogue on Covid-19 — she philosophises on the nature of numbers, the way we overestimate their truthfulness and underestimate their power to shape our thoughts and view on reality and how we should hold them to moral standards. She sparks her arguments with tons of examples from American life. She brings in anecdotes from politics, health, employment, fitbits, crime profiling… and while some will undoubtedly inspire, others may feel odd and even alien.
Numbers, she claims, strip away valuable data when you use them. After all, counting starts with deciding which things are “alike”. You can count apples and pears when you abstract them to ‘pieces of fruit’ but in the process you lose the distinction between the apple and the pear. Actually, you already entered the problem when you named the Granny Smith, the Golden as well as the Pink Lady ‘apples’. And you most certainly did not mention how sour the first is, how old and wrinkled the second one and that someone bit in the third.
Of course we don’t care about the apple’s feelings but when it comes to counting people the nuances, the differences, the diversity that is lost when we select the traits with which we want to count people can be problematic. She addresses the innate judgment that lays in that selection. When we discern age groups we have already judged that age is a relevant and valid distinction. When we group people according to a vague notion like race we give importance to it and consolidate the distinction, creating a harsher and more clean cut reality than was there before we started counting. Counting starts with the preconception of what matters and what doesn’t, who ‘counts’ and who doesn’t.
She stresses that when we talk about numbers and data later on in a seemingly objective mathematical way, we tend to forget that the choices we made when we started counting, were very subjective and carry our prejudices and unchallenged, often messy patterns.
Counting is shaping
“Counting changes our relationship to the world”, the author says, “because in order to count, we narrow our focus to one or a few features of whatever we’re counting. […] While we count them, we stop noticing their differences. […] Counting heightens our awareness of the things we’ve decided to count and makes us ignore things we (or someone else) decided don’t matter.” In this way she claims that coming is a power play. You force things into categories by ignoring the differences.
“When polls and surveys attempt to read people’s minds, they also change people’s minds”, she continues giving an example from the American National Election Studies (ANES) that wants to inquire into the people’s prejudices with questions like “Do you think that most white candidates who run for political office are better suited in terms of their intelligence to serve as an elected official than are most black candidates, that most black candidates are better suited in terms of their intelligence to serve as an elected official than are most white candidates, or do you think white and black candidates are equally suited in terms of their intelligence to serve as an elected official?”
Stone argues that these kind of questions teach that generalizing about a group like this is an acceptable way to think, even though they conjure derogatory images. “If the questions were parents, I’d tell them they’re modeling bad behavior to their children”, she adds.
Naughty maths? Not really
The author keeps hammering on the same nail, wanting to convince us about how there is life beyond the numbers and that numbers lose those lively details. Of course she stipulates that statistics can be used to manipulate and misrepresent an underlying reality. And while she is not completely wrong, when she points that out, she does seem to bash maths a bit too harshly.
Most data analysts and data imagineers know the problem of the loss of detail and the importance of choosing the right categories so well, that a huge part of data gathering and data management is exactly about addressing that. The meticulous search for what matters is the reason why our statistics are as comprehensive and often complex as they are. Higher-level math involves exactly the complexities she is asking us to be aware of.
During a large part of the book I was struggling with the uneasy feeling that her abstraction of what statistics bring to the table is suffering from the same kind of prejudice she rightfully wants to warn us for. She too seems to leave out the important ways in which data can enlighten and help us without becoming a fiction as she repeats all too often.
Humane AI
Data, data gathering, data processing, data analysis does not have to feel as cold and disconnected as Stone depicts it. Even with added Machine Learning and Artificial Intelligence data does not have to lose its human foundations. But yes, we need to inject consciously inject the human touch back in there. When someone comes into a store day after day after day buying boxes of liquor, the machine thing would be to sell him or her even more, pointing to even more expensive drinks. A more Humane AI would see the pattern for what it is and try to address the call for help in there. And just as Humane AI wants to aid humans in key human goals instead of exploiting them for profit, data too can be Humane, bringing the attention back to the life beyond the numbers. It addresses the issue Deborah Stone points out. Brené Brown calls stories “data with a soul”. Humane AI, Humane Data Analysis or Data Imagineering reveals the stories hidden in the numbers and the graphs , or, as Stone puts it: it puts “the soul back into numbers “.
Metaphors, Tipping Points and Blind Spots
It is important to remember that all data starts with that creation of categories. And that feeling kept persisting in my mind throughout the book. I was reminded of the famous Korzybski quote “The map is not the territory” where he too warned for mistaking words or categories for the thing itself. All presentation is metaphoric after all. We deal with metaphors, we think in metaphors… and while we are extremely good at it, it does pay to question our metaphors from time to time: do they not create new blind spots because we missed a distinction, because we skipped an important detail when defining our categories?
Maybe the tipping point we are looking for in our standard numbers and data can not be found because we have been looking at the underlying world in a wrong way? We often use demographic data to categorise people but maybe we need new categories to find the right answer instead of asking the old questions again and again? Maybe the tipping points can’t be found in the data because we overlooked them from the get-go.
With The Data Arena we do want to go back to the basics of the data, to the world beyond the numbers and the decisions that went into the collection of them. How can we unearth what we don’t know yet, even though it is so very important? How can you find the blind spots and make them visible? How can we tell which ones may be groundbreaking? How can we find the real triggers, the golden nuggets of data? How do we find the finest ways to categorise our reality to attain the most usable, valid, reliable and effective data?
To read or not to read
No, we did not find the answers to these questions in this book. We didn’t even find the questions in it. But it feels like they need to be addressed.
And even though you cannot fault a book for having another purpose than the one you are interested in, we would not catalog this book as a ‘must read’ unless you are very invested in US public policy or the potential misuse of data. The style is awesome, Deborah Stone sure is masterful storyteller, and though we do agree numbers need a human touch we rolled our eyes at her overly negative attitude towards data and a mathematical approach of knowledge.
Counting: How We Use Numbers to Decide What Matters. Deborah Stone. Liveright, (288p) ISBN 978–1–63149–592–2
