Rebel Ideas- the Power of Diverse Thinking

by Matthew Syed

  Daniels’s work led to a dawning realisation. A cabin standardised to the average pilot may sound logical, even scientific, but it is fraught with latent danger. The standardised cockpit was the root cause of the alarming incident rate, causing multiple crashes. And it forced the Air Force to think in a new way about design. Instead of requiring the pilot to conform to a standard cockpit, which suited almost nobody, they re-designed the cockpit to adapt to the diversity of individuals.

  Sure enough, when planes were designed to enable pilots to vary the height of the seat, the distance of the joystick, and so on, incidents plummeted. Moreover, the cost of creating this flexibility was minimal compared with the savings on incidents, not to mention the human cost of injuries and fatalities. In turn, the safety performance of the US Air Force soared.

  III

  The standardised cockpit of the US Air Force was not just a hazard, it is also a metaphor. It is just one example of the standardisation of our world. We have standardised education, standardised working arrangements, standardised policies, standardised medicine, even standardised psychological theories. All, in their different ways, fail to take into account human diversity. They treat all people as manifestations of a mythical average rather than as individuals. This takes us back to the point at the beginning of this chapter about how this flaw can cause us to overlook human diversity, and lose its benefits. We are all different from one another. We have different physical dimensions, but also different cognitive traits, strengths and weaknesses, experiences and interests. Indeed, this is one of the most wonderful things about our species.

  But if we differ in important ways, enlightened systems should, where possible, take account of this variation. Indeed, we should celebrate it. After all, how can we reap the benefits of human differences when we are crowbarred into rigid systems (and not just cockpits)? How can we harness diversity when we are deluded by averages that obscure the ways in which we differ from each other?

  Let us take a brief example to highlight the logic. In an experiment at Google in 2014 a team of psychologists gave a short workshop to staff in sales and administration. Such work tends to be performed in a standardised way, at the same times, and with the same tools. This standardisation is not physical, but conceptual. Indeed, the idea of building flexibility into such jobs seemed like madness. After all, these are the admin and sales people, not the whizzy engineers.7

  But the workshop encouraged the professionals to think of their jobs not as fixed parameters, like inflexible cockpits, but as adjustable designs. They were taught to consider how they could play to their strengths, shaping the contours of their work around their interests and talents, as well as the objectives of the company. They were asked, in effect, to think of themselves as individuals with distinctive skills and insights rather than homogenous cogs in a machine. As Adam Grant, one of the co-researchers, put it:

  We introduced hundreds of employees to the notion that jobs are not static sculptures, but flexible building blocks. We gave them examples of people becoming the architects of their own jobs, customising their tasks and relationships to better align with their interests, skills and values – like an artistic salesperson volunteering to design a new logo and an outgoing financial analyst communicating with clients using video chat instead of email . . . They set out to create a new vision of their roles that was more ideal but still realistic.8

  What happened? Those who attended the workshop were rated by managers and co-workers as happier and higher performing, and were 70 per cent more likely to land a promotion or move into a preferred job when compared to a control group. Grant writes: ‘Instead of using only their existing talents, they took the initiative to develop new capabilities that enabled them to create an original, personalised job. They became happier and more effective – and qualified themselves for roles that were a better fit.’

  We noted in Chapter 3 that taking averages can be effective in forecasting. You’ll remember that the average forecast of six economists was significantly more accurate than the forecast of the top economist. In this chapter, however, we seem to be arguing that averages are bad. How do we make sense of this difference? Is there a contradiction in the analysis? In fact, these two perspectives are not just compatible but complementary. The economic forecasters had different models. They expressed their estimates independently. They were free to come up with their own predictions. Taking the average of these different perspectives was a way of aggregating diverse information while filtering out the errors.

  Standardisation is different. This is where people of different sizes are forced to inhabit the same cockpit. Or where people are asked to do their jobs in the same way, regardless of their differences, squeezing diversity out of the picture before it has even had a chance to manifest itself. It would be rather like forcing economists to use the same model; the model used by the average economist. This would effectively eliminate useful differences. Rebel ideas would dry up.

  To put it another way, averaging diverse predictions is a way of exploiting diversity. Standardising the way people work, or learn, or whatever else, risks squashing diversity. As Neil Lawrence, head of machine learning at Amazon, puts it: ‘When an average is being used well, it’s harnessing the insights from multiple people. When it’s used badly, it’s imposing a solution for multiple people.’9

  Of course, standardisation can often be useful and valuable. With clothing, for example, off-the-shelf options may not always fit perfectly, but they enable consumers to gain the benefits of cheap, mass-produced apparel. Individualised solutions (made to measure) are typically more expensive, implying a trade-off between the bespoke and the generic. But often the generic is adopted not because it is more cost-effective but because we scarcely consider the alternative. That was the case with cockpits, where standardised designs were created not after a cost-benefit analysis but because few conceived of the possibility that a cockpit designed for the average pilot might not suit most pilots – at least until Lieutenant Daniels came long.

  When institutions are too rigid, everyone suffers. This is true not just of organisations, but the way in which patterns of thought that orbit the concept of average cause us to overlook human diversity in subtle ways; patterns so deep that they have infiltrated science itself. With this in mind, let us return to Eran Segal. For this will illuminate the risks – and why they extend far beyond the question of what we put in our stomachs.

  *

  After his low-carb marathon run, Eran Segal was finally zooming in on the flaw in dietary science. Standardised dietary guidelines, like standardised cockpits, might seem rigorous, but they overlook a key variable: the diversity of people. Segal says:

  A good example is the so-called glycaemic index. This is a system of ranking foods according to how much they influence blood sugar. The way to obtain such an index is to take a group of people, get them to eat different foods, and then measure the response. This way you can obtain an index ranked from one to one hundred which rates food accordingly.10

  Described in this way, the glycaemic index sounds like a gold standard of science. It is built around measurement and data. But it is also built around something else: the average response to food. But what if people react to the same food in fundamentally different ways? People who base what they eat on the glycaemic index might be eating food that is, for them, unhealthy. In the spring of 2017, Segal and his fellow researchers conducted an experiment to test this possibility. The objective was to measure the response of subjects to two different types of bread: the commercially produced, white bread often demonised by the health lobby, and the handmade, whole-grain sourdough beloved by health nuts. As ever, the existing evidence was mixed. Some studies suggested that bread could reduce the risk of cancer, cardiovascular disease and type 2 diabetes. Other studies suggested pretty much the opposite; namely, that bread had little effect on clinical markers of health.

  Segal’s experiment was simplicity itself. He took a group of healthy people, n
one of whom were on any particular diet. They were then randomly assigned to two different groups. Some ate white bread every day for a week, while the others ate brown. Neither group was permitted to eat any other wheat products, and were required to eat only bread for breakfast, while including it in other meals as desired. The two groups then took a two-week break, and then switched to the other diet.

  Crucially, every person in the experiment was measured multiple times for how they were responding to the bread. Various measures were tracked including inflammation response, nutrient absorption and more. Perhaps the most important measurement was the blood sugar response. This is crucial to health, and it is worth spending a moment or two to explain why. One of the first things biology students learn about the human body is the importance of glucose metabolism. After we eat, our body digests the carbohydrates, breaks them down into simple sugars, and releases them into the bloodstream. From that point, with the help of insulin, glucose is moved into the cells and liver, where it is used to synthesise glycogen for later use as energy.

  However, insulin also signals cells to convert excess sugar into fat and store it – this is a primary reason for weight gain. Also, if too much glucose flows into the blood from food, it may cause an over-production of insulin, pushing glucose levels too low. This makes us hungry and keen to eat more, even though we have had more than enough food. Sharp glucose spikes are a risk factor for diabetes, obesity, cardiovascular disease and other metabolic disorders. One study that followed two thousand people for more than thirty years found that higher glucose responses predicted higher mortality. Chronically high blood sugar levels put stress on the whole system. Steady blood sugar, on the other hand, with modest and gentle rises after eating, may reduce heart disease, cancer and other chronic diseases such as excess fat, and mortality. In short, the blood sugar response is important not just in terms of weight, but health, too.

  When the results from Segal’s bread experiment came rolling in, it turned out that the two different breads made no difference when it came to blood-sugar response or any of the other clinical markers. Industrialised white bread and handmade sourdough had virtually the same effect. This seemed to imply that dietary advice should be neutral. If one bread is no better than the other, why not advise consumers to select the one that tastes better, or is cheaper?

  And yet this ‘scientific’ inference was based upon the average response. What about the individual responses? Were people diverse in the way they responded? The results were remarkable. Some people showed benefits from eating the whole-grain sourdough and adverse effects from the commercial white, while others had the opposite response. Some showed little difference between the two, while others showed dramatic differences. ‘The whole data set was highly personal,’ Segal says. ‘You had to look at the individuals, not just the averages.’

  Why were the responses so different? Segal realised that in somewhat the same way that body size has multiple dimensions that influence whether a person will fit into a cockpit, the human body has multiple dimensions that influence how an individual will respond to a given meal. These dimensions include such intuitive things as age, genetics, lifestyle and more.

  Perhaps the most fascinating dimension is the microbiome, the bacteria we all host in our gastrointestinal systems. There are around forty trillion cells and up to a thousand different microbial species in our bodies. This ‘universe’ has about two hundred times more genes than the host human, and exerts a major influence on how we digest food and extract nutrients, as well as on the immune system. And these microbiomes vary from person to person.

  When you look at diet from this perspective, with different factors translating into different enzymes, genes, bacterial genes, and perhaps dozens of other unique factors, it seems almost absurd to suppose that any diet could be sensible for all, or even most, people. ‘The more I thought about it, the more curious it all seemed,’ Segal says. ‘Standardised dietary advice will always be flawed, because it only takes into account the food, and not the person eating it.’

  Perhaps Segal’s most ambitious study went further. Many experiments take a small group of people, give them some treatment or intervention, and then measure the average impact at a specific moment in time. Segal’s experiment recruited almost a thousand subjects. Around half of these were overweight and a quarter were obese, matching the non-diabetic population of the developed world. These subjects were then connected to a glucose sensor and tracked every five minutes for an entire week, resulting in individualised blood sugar responses for almost fifty thousand meals.11

  The subjects logged everything they ate on a specially designed mobile app. They were allowed to eat what they wished, but they were required to eat a standardised breakfast: a rotating menu of plain bread, bread with butter, fructose powder mixed with water, or glucose mixed with water. This created a rich data set, including a total of 46,898 real-life meals and 5,107 standardised meals, with 10 million calories logged, along with associated health data. For an experiment involving nutrition, this was on a different scale to almost anything before. And instead of calculating an average response, Segal and his colleagues examined the response of every individual.

  The results, when they came in, were stunning. For some people, eating ice cream led to a healthy blood sugar response, while sushi had the contrary effect. For others, it was the other way around. ‘For every medical or nutritional finding that came up, there were many people whose results were very different from it,’ Segal said. ‘Often people responded in diametrically opposed ways.’

  Keren – Segal’s wife – was staggered by the results. A trained clinical dietician, she had seen dozens of patients in her clinic and relied on general guidelines to offer advice. With pre-diabetics, the advice is to stop eating ice cream and switch to complex carbs, like rice, instead. ‘I realised that I had been giving people advice that could harm them,’ she said. ‘It was sobering. Now I advise them to measure their own blood sugar responses. That way, they get a diet that works for them.’12

  Talya is an archetypal example. A sixty-four-year-old retired paediatric nurse from northern Israel, she was heading towards diabetes. She was clinically obese, and increasingly worried about her health. ‘I was putting on a lot of weight,’ she told me. ‘My blood sugar levels were very high.’ She was eating in a seemingly healthy way: omelette for breakfast, balanced meals through the day, and plenty of fresh fruit and vegetables. She grew her own produce in her backyard and particularly enjoyed apples and nectarines. ‘It seemed to be as good a diet as I could manage,’ she said. ‘I couldn’t really figure out what I was doing wrong.’

  When she was given a glucose sensor so she could take regular measurements of her personal response to meals, she was dumbfounded. She spiked for nectarines, melon and tomatoes. She also spiked for milk with 1 per cent fat. Yet her blood sugar response was perfectly healthy for watermelon and 3 per cent milk. ‘It was astounding,’ she said. ‘I had no idea this was happening.’

  Talya altered her diet to match up with personalised guidelines, losing 17 kilos and lowering her blood sugar levels by 20 per cent. ‘No two people are exactly the same,’ she said. ‘We have different DNA, different biology. I am married to a man who is very skinny. Before, when we ate the same things, his blood sugar was fine. Now my blood sugar is coming down to normal . . . Who on earth would have guessed that I would have a problem with nectarines!’

  But Segal’s study was not yet finished. The researchers then pulled all the data into an algorithm designed to predict blood sugar responses for new participants. Effectively, they were using a similar approach to the way online retailers like Amazon predict the kinds of books that shoppers will like. To test the algorithm, one hundred new people were recruited and then measured on personal characteristics such as blood, age, microbiome and the like. This data was then fed into the algorithm. This was a significant test of the research. Would the algorithm more accurately predict how people would respond to differen
t meals than standard carbohydrate counting?

  The answer was an emphatic ‘yes’. ‘It was a huge thrill to see that we could take any person, even people who were not part of the original study, and predict their personalised glucose response to any meal with good accuracy,’ Segal said. ‘It gave us assurance that the algorithm was robust.’

  Finally, they recruited twenty-six new participants with pre-diabetes and asked the algorithm to design two diets for each subject. In the ‘good diet’, the algorithm was asked to predict meals that would have a low blood sugar response. In the ‘bad diet’, it was asked to predict meals with a high response.

  By now, you won’t be surprised to hear that the ‘bad’ diet for some was similar to the ‘good’ diet for others. One person’s good diet was composed of eggs and bread, hummus and pita, edamame, vegetable noodles and tofu, and ice cream, while their bad diet was composed of muesli, sushi, marzipan candy, corn and nuts, chocolate and coffee. As predicted, the ‘bad diets’ were associated with abnormally high glucose levels, and impaired glucose metabolism. On the good diet, despite the same number of calories, glucose levels remained completely normal, without a single spike across the entire week. ‘These results were frankly stunning to us,’ Segal said. ‘It was proof that you can manipulate your own blood sugar levels so significantly that you can go from pre-diabetic blood sugar levels to normal in one week, only by changing your food choices.’

  These results are important in their own right, but the key point – for our purposes – is that that they deepen our comprehension of diversity. Presuming that pilots all conformed to the dimensions of the average pilot led to a litany of incidents in the early 1950s. The same conceptual flaw has persisted, almost unnoticed, at the heart of nutritional science. Unless you take into account the diversity of individuals, you are likely to design systems, guidelines and much else that are defective or restricting or both.