Are we smart enough, Gerd?

A friend of mine once remarked that “We are not smart enough for our age [epoch]”. What could they have meant by this? Education levels across the world are remarkably high, and many believe ourselves capable of good decision making. Yet, many aspects of the world around us elude, deceive, and confuse us. From time to time we find that it is exactly those who should know, who are falling short of expectations. In these moments, that glossy shine of modernity, brittles away so slightly, and we feel fooled.

No doubt, there is naturally opposition to this broad statement — one would say that, No! we are in fact smart enough for our age, we have mastery over the skies and the earth, our advancement in technology is not comparable to any epoch before us. It is true, surely, us, taken as a whole, are collectively smart enough; and who could be surprised, as much of this world has been established by us. Indeed, we do know how electricity works, we do know about carbon dating, we can perform heart surgery. And yet, why do surveys find that most doctors do not know how to weigh the benefits and dangers of simple medical screens, this ought to be their expertise? To resolve this butting of heads, the we here refers to sum of individuals navigating life as independent beings, as opposed to the totality of the human species and all of its information and knowledge. We can simultaneously know how to build atomic weaponry, whilst having the majority of our world be statistically illiterate. The two are mostly unrelated. On one hand the repository of all human knowledge, the other, the understanding embodied in a single person.

Maybe, “we as people, are not smart enough for this age”. The social scientist Gerd Gigerenzer might say, “We as people, are not risk literate enough for this age”. The way I take it then, is that my friend was discussing us as people encountering world, living our lives in modern society, with all it’s technological features. Gerd, in his book Risk Savvy, opens by touching on something similar;

Literacy—the ability to read and write—is the lifeblood of an informed citizenship in a democracy. But knowing how to read and write isn’t enough. Risk literacy is the basic knowledge required to deal with a modern technology society. The breakneck speed of technological innovation will make risk literacy as indespensable in the twenty-first century as reading and writing were in previous centuries. Without it, you jeopardize your health and money, or may be manipulated into unrealistic fears and hopes. One might think that the basics of risk literacy are already being taught. Yet you will look in vain for it in most high schools, law schools, medical school, and beyond. As a result, most of us are risk illiterate.

I feel that this is what we’re getting at here. Not to pile upon our inadequacies—but put in another way— we are not adjusted, adapted enough to our age. We are still playing catchup, running, trying to keep up with the pace of it all, the speed and volume of information, and the seeming legitimacy that it holds. Our tools, techniques and methods to navigate and interpret aspects of our time, are simply not developed sufficiently.

As moderns, I don’t think it misguided to say that we excel in putting forth predictions, and specifically, quantifying those predictions. Through our technologies of measurement, and our analytical methods, presenting probabilities of events are a regular, if not, default characteristic of our communication. Perhaps we could go so far as to say that they are an expectation at this stage. At our doctors offices we seek likelihoods of x, we hear weather-women telling us of rains and droughts, financial analysts the trends in exchange rates for next year. We are accustomed to think numerically and so the allegations of illiteracy do strike us unjustified. Despite this numeracy, how often do we take the next step—putting confidence values on those predictions themselves—and—placing them under scrutiny. They are taken for granted, as concrete as heat from fire. Not all numbers are created equally, unfortunately.

Looking at some examples, Gerd describes an interactive lecture that he presented to a room of gynecologists at a medical networking event. In Australia, the time it takes to receive a qualification in order to practice gynecology is 10 years at minimum, with 12-14 years being more common. These are highly qualified individuals, perhaps the most so in our societies. He presented the room with a question; “If a women undergoes a routine mammography screening, and tests positive, what is the likelihood that she actually has breast cancer?”. He proceeded to provide the rates of breast cancer among women, and the test efficacy statistics; rates of true to false positives and negatives. Collecting the doctors answers via an interactive polling system, he found that only 21 percent of the doctors would correctly inform women of their likelihood of having breast cancer given a positive screening. One has to wonder what could be going wrong here. Much of Gerd’s work is dedicated to understanding why these experts are not as literate as they should be, and fortunately, many of his answers provide redemption. These same test efficacy statistics, when transformed into natural frequencies, often allow people to solve these same problems which were once too complicated when presented in a classical probabilistic form¹²³. This could be as simple as transforming a percentage to an easily parsed ratio e.g. 1% to 1 in 100, and performing a rudimentary calculation on the input values, allowing the answer to reveal itself far more obviously.

Even then, leaving it in doctors hands doesn’t sit well, personally. Our technological sophistication has led to high degrees of domain specific knowledge which are acquired through these years of education. It does feel as though everybody is represented as an expert at something. Some of these disciplines are defined to a degree by their mandate to make these forecasts. A sort of monopoly of legitimate prediction, ala, the state and it’s violence. They are provided such social mandates because they are experts (note the dangers of this cyclicality if the quality of experts is compromised). As such, it is entirely normal to put our trust in them, and a trust that is held in high regard is a pillar for social cohesion. There is a really sinister side of this illiteracy—there are real people who bear the consequences of misdiagnosis—bodies are mutilated, made incontinent, subjected to hopeless chemotherapeutic treatments with troubling side effects, tremendous resources laid to waste along the way. These are all features of a technological age which should know better, which will know better. Gerd’s insistence is that we should no longer view this as a hole that even more domain knowledge is needed to fill, but the beginning of general, broad sweeping risk and statistical literacy from the base up. I was shocked to read that sixth graders’ performance with natural frequencies matched the performance of adults with probabilities graders⁴. The organisation of information matters, not just it’s existence!

These shortcomings become even more urgent, and are compounded by the intentional misrepresentation of information; naked deception. Across numerous fields, not just health and medicine, there are bodies who purposefully misconstrue for their own gain. A conscious propagation of misunderstanding, most often for monetary benefit and gain. Opening up a webpage, our University is boasting of a research group that has found a 20% increase in survival, which, standing on it’s own, seems remarkable. That is until the absolute increase is revealed. Survival has increased from 1 in 10'000 to 1.2 in 10'000. Would you get on a treatment for a disease which increased your survival chances from 1 to 1.2 in 10'000? To hammer a previous point, let us not forget the dark flip side of these interventions, the side effects. Thinking both ways, in both directions, is a feature of risk literacy.

Living in freezing environments necessitates knowledge about how to heat ourselves and our homes, when it is safe to move about, how we can feed ourselves. Surely we don’t want to just survive, but thrive too, and live good lives. While our technology rich societies may not pose immediate dangers to our lives the way the deep cold does, it is not so different here. Do not forgot, just as there are those who die in the freezing wilderness for lack of preparation and understanding of the climate, there are those who have also perished due to their lack of understanding of this dimension of our new environments. Unfortunately, it is often those who bring us into the blizzard, that are convinced we are going to oasis. And so, living in an environment, a world that is flooding, and steeped in information represented in statistical and numerical form, it becomes an imperative to acquire an understanding of how to live well in this world; and just as we ignore the chill of the cold on a freezing night, when we ignore probabilities and risk literacy, it is as our own peril. How often do those who are triggered by online rage bait, reflect on whether they are being hoodwinked? Trapped in an artificial, algorithmically generated pocket of reality where the entire purpose is to present the events in question as being more important than they likely are, to amplify the probabilities of their actual existence. One cannot fault us entirely for our misreadings, it is not often in physical, blood-and-bone reality that we see an object or environment, and have to question what the likelihood that this environment actually exists is — the object and environment are given, we see, smell, touch them.

A second type of response that’s often put forward to this notion that ‘we are not smart enough for our age’ goes something like “No, we are smart enough, we have the tools, and this is exactly how you solve it, this is the optimal solution to the problem.” While the first response defers to expertise, which, paradoxically, doesn’t exclude experts themselves who actually defer to their own expertise, the second response often assumes the position of the expert themselves, and defers to specific techniques and technologies. According to Gerd, two common approaches are encountered; a) newer is better, and b) optimisation with more information. In the ’newer is better’ approach, the solution to a problem is sought by finding the most sophisticated and thus typically newest method of answering it. Here sophistication is linked to recency, and recency to effectiveness.

An example; acute vestibular syndrome typically manifests itself with dizziness, nausea and vomiting. When a patient enters the ER, doctors need to quickly rule out whether the patient has had a stroke, which in rare cases causes this syndrome. The doctors can either send the patient for an MRI, which takes 15-90 minutes and costs over $1000, or, as skilled doctors had done for decades, undertake a bedside exam known as HINTS. HINTS is a quick manual evaluation where eye tracking and visual responsiveness is used as a reliable proxy for whether the patient has this syndrome. In a study of 101 high-risk patients with acute vestibular syndrome, diagnosis was mode with both an MRI and the HINTS exam. While MRI missed strokes in 8 out of 76 patients, the HINTS examination caught all correctly and did not miss one. The exam caused one false alarm amongst the patients without stroke. In total, the HINTS exam was far quicker, far cheaper, far more comfortable for the patients, and most importantly, it was in fact more accurate than the multimillion dollar MRI device. To bang the drum once more, as this is one of Gerd’s most committed points; there are methods which are fast, frugal, relying on minimal information, yet, when compared to shinier methods, actually outperform them.

Being risk literate involves knowing when to use certain techniques and tools over others, and not being seduced by the latest developments for the sake of novelty. In a climate of rapid change and acceleration, this can be a trying demand.

Optimisation with increased information is a roundabout way of saying that we add more data into an equation, or set of equations, and compute the optimal solution to said problem. These types of problems are often found within the realms of economics and engineering, but there are notable other examples. In Risk Savvy, Gerd describes a common scenario for those looking to invest some of their money in a diverse way and to avoid putting all their eggs in one basket. A common goal here is to minimize risk for a given return on ones investment. Harry Markowitz won a Novel Prize in economics for solving this problem, demonstrating that there is an optimal way to allocate ones funds — a mathematically correct solution. This and similar methods are employed by banking and financial firms for their investments. Where the story gets peculiar, is that when asked about his investment retirement strategy, Markowitz decided to use a simple rule of thumb, the 1/N method. Split your investments equally. Love all your children the same. When asked why he didn’t use his Nobel method, he replied by saying that he didn’t want to be embarrassed if the stock market acted differently than to what he expected! Why the caution from the one who outlined an optimal strategy? When 1/N was compared against Markowitz’s own, and a dozen other complex methods in seven real world situations, such as picking ten American industry funds, 1/N outperformed all of the methods in 6 of the 7 situations, as judged by common performance criteria⁵.

A lastly notable example from the book; we go to a restaurant for the first time, without any background information about it. What do we order? Is there an optimal answer to this problem, one where we will maximize our satisfaction and find the best item on the menu for us? Let’s go pursuing this optimal solution, off we go, creating a decision matrix from all the items, assigning scores to each item, then calculating a maximum pleasure score for the one we should choose. Or… should we order all the items and taste eat one fractionally? Well, perhaps now, in the realm of LLMs, such data compression is possible, so we pull our phone out and ask our chatbot. I hope it’s not surprising to say that all these methods are frankly a bit ludicrous, no? Even pulling out our phone and trawling the chat engine has a smell of the pathetic to it. The quickest, and likely most effective solution, is just asking the waiter what the crowd favorites are, and choosing amongst these. Assuming we’re at a respectable establishment that isn’t trying to sell you their almost-off steaks, you’re likely to leave happy. I like this example as it magnifies certain core restrictions, and limits on our everyday life — time is not infinite, computing power is not infinite, wealth is not infinite, we cannot sample infinitely. Effective decision making is both accurate, that is, get’s close to the correct answer, and is also aligned to the real, living situation at hand. There is far more than simply correctness or optimality, but speed, repeatability, clarity, simplicity — the chase for optimality may thus obscure actual decision making. Perhaps with all the dishes at hand, and a year to waste, we could find what we best enjoy, but then on the first of January, a new head chef would arrive, and we’d be back to square one.

This quest for optimality, and this bias for optimisation as a default problem solving method, has additional pertinence in the present moment, where AI hype is ballooning expectations in many facets of our lives. It is often believed that answers to a problem are all found at the end of a very serious calculation, with shortcomings stemming from inadequate data, rather than misguided method. Here we have manifested an insatiable appetite: more compute, more parameters, more data, more electricity. One would be led to the thought that the human mind was a 200sq metre data center. Let’s not charter into the worlds where optimal solutions may not be computable at all. A moments reflection follows thereafter.

Gerd, in Risk Savvy has distilled the importance of understanding ones domain, and viewing it in much the same way as an environment straight out of “nature” itself, with qualities and attributes which ought to be taken as seriously as a strong gust of wind. Thanks to the chasm of time passed by our ancestors, we have adaptive strategies which extremely effectively solve natural problems, and well, time moves forwards, and we need new strategies to add to our adaptive toolkits, for today and tomorrow. We have remarkably powerful tools and techniques, but, we need the literacy to know their limits, and the honesty to confront hubris and overconfidence.

Gerd might say, we are NOT smart enough, but we can be savvy enough!

No doubt, this has implications for democracy. To wrestle the power from those who believe us best managed by small groups of elites, hereditary or otherwise. An illustrative example being Galton’s ox weighing⁶ - maybe the topic of another scribble.