It all started when Robert F. Kennedy Jr., our Health Secretary, rolled out new dietary guidelines this year, aiming to “Make America Healthy Again.” Now, you’d think something like that would be met with universal applause, right? Not quite. It was more like a mixed bag of reactions, a bit like a potluck dinner where some dishes are a hit and others, well, not so much. Some folks, like the American Heart Association, actually gave a hearty cheer for the renewed focus on veggies, fruits, and whole grains. They were all for it, happy to see these familiar healthy foods getting the spotlight. But then there were others, who raised an eyebrow, especially at the promotion of red meat and full-fat dairy. They even went as far as to accuse Kennedy of spreading what they called “blatant misinformation,” claiming that “healthy fats” included things like butter and beef tallow. Imagine the uproar, the headlines, the passionate debates across dinner tables and social media feeds! It was a real whirlwind of differing opinions, highlighting just how complex and deeply personal our relationship with food and health truly is.
This whole debate, especially the use of the word “misinformation,” got me thinking. It’s a word we hear everywhere these days, in the news, in casual conversations, and it’s often used with a lot of weight. And sometimes, for good reason! When genuine lies are spread, they can really mess things up – democratic processes can be undermined, people’s health can be harmed, and sometimes, it can even stir up violence. As someone who works in AI strategy at the University of Waterloo, I’ve seen firsthand how worried people are about artificial intelligence making this problem even worse, amplifying these untruths across the digital landscape at a speed we can barely comprehend. But here’s the thing: “misinformation” is also a loaded word, isn’t it? It feels like there’s a growing trend where people slap that label on just about anything they disagree with, even if it’s not a genuine lie. It’s almost as if once you say “misinformation,” the conversation is over, the other person is wrong, and there’s no room for nuanced discussion. As a professor of statistics, I believe a big part of this problem stems from how incredibly difficult it is for us, as humans, to truly assess evidence. It’s not always black and white, and understanding what constitutes “evidence” is far more complex than most of us realize, leading to a lot of this conversational deadlock.
Let’s dig into that difficulty of assessing evidence for a moment, because it’s a crucial point. When someone declares, “there is no evidence that eating red meat is harmful,” or conversely, “there is evidence that full-fat dairy is bad for your health,” it sounds pretty definitive, doesn’t it? But actually, substantiating these kinds of statements is incredibly tricky. Part of the challenge is that it’s often incredibly hard – though not entirely impossible with advanced statistical wizardry – to pinpoint the exact effect of one specific habit. Think about it: our health is a complex tapestry woven from our genetics, our environment, our overall lifestyle, and countless other entangled factors. Trying to isolate the impact of, say, eating red meat, from everything else going on in a person’s life is like trying to untangle a single thread from a very complicated knot. This is why you’ll often see research studies carefully stating that they’ve found an “association” or a “correlation” between food consumption and health effects, rather than a direct cause-and-effect. They’re being precise because it’s genuinely hard to prove causality. But even in situations where you’d think it would be crystal clear, where there are no obvious entanglements, assessing evidence remains surprisingly difficult. Imagine for a second, you’re playing a game, and your opponent rolls a die seven times. Six out of those seven times, it lands on an odd number (one, three, or five). Now, in theory, odd and even outcomes should be equally likely, right? So, this seems a bit off. The question then becomes: is this skewed outcome evidence that the die might be loaded? Does it suggest someone might be cheating? This simple scenario beautifully illustrates the everyday challenge we face in interpreting data and deciding what constitutes “proof.”
Now, let’s play with that loaded die scenario a bit more, because it really highlights how different ways of looking at evidence can lead to seemingly contradictory conclusions. If you were to use a very common statistical tool called the “p-value,” you might actually conclude “no,” the die isn’t necessarily loaded. The reasoning here is that there’s still a pretty decent chance for a normal, fair die to show an odd outcome more than five times out of seven rolls. So, rolling six odd numbers, while a bit unusual, isn’t as shockingly unexpected as it might first appear to that particular statistical framework. However, if you switch gears and use a different evidential scale, known as the “e-value,” you could argue “yes,” the die probably is loaded. Why? Because a loaded die would be much, much more likely to produce six odd outcomes out of seven rolls than an unbiased one. So, from the e-value’s perspective, those six odd numbers are much more consistent with the suspicion that the die is rigged. It’s almost like having two different sets of glasses – one makes the world look one way, the other a slightly different way. This brings up a critical point: how can two opposite arguments both be correct? Or can they? The truth is, both arguments need an implicit “threshold” to reach their “yes” or “no” conclusions. For the p-value, the threshold is about how small a probability needs to be to be considered “unlikely.” For the e-value, it’s about how much “more likely” an outcome is under one hypothesis than another. These thresholds are often arbitrary, and because they’re different, one method might end up saying “black” and the other “white,” when in reality, the situation is just a certain shade of grey. Statisticians can actually calibrate these decision thresholds so they always reach the same conclusion, but for the average person, we’re largely unaware of these nuances, and the tools we use unconsciously shape our conclusions.
This brings us to a crucial point about human psychology and how we react to information, especially when those underlying scales and thresholds change without us realizing it. The average person isn’t very good at performing this kind of statistical calibration intuitively. Our brains tend to react very differently depending on how information is framed. Imagine you’re told that people who regularly eat a certain delicacy are 25 times more likely to develop cancer later in life. That’s a pretty startling number, isn’t it? Your gut reaction might be to immediately ditch that delicacy. But what if the information was presented differently? What if you were told that eating it would increase your probability of cancer from a tiny 0.01 percent to a still relatively small 0.25 percent? Suddenly, “25 times more likely” might not seem quite so terrifying. You might decide, “Well, 0.25% is still a very small risk, so I’ll keep enjoying my treat.” Neither of these choices is inherently right or wrong; they are personal decisions based on how we interpret and weigh risk. And this is where the “misinformation” problem becomes truly exasperating in our social discourse. If I were to say today, “You know what? Given those risks, I don’t feel the need to change my diet,” I’m afraid that people who are super enthusiastic about changing theirs might gang up and accuse me of spreading “misinformation.” This kind of immediate, unthinking labeling has to stop, before it completely poisons our ability to have meaningful conversations. We need to reserve the word “misinformation” for genuine, deliberate lies, not for conclusions that simply stem from subjective thresholds or different interpretations of complex evidence, even if those interpretations are common. Declaring something to be “evidence” or “not evidence” simply because a p-value is above or below some conventional line has already led to far too many research findings that can’t be replicated, casting a shadow on scientific progress itself. To hurl the accusation of “misinformation” based on such shaky evidence, or even a lack thereof, does nothing but impede true scientific discovery and honest, open dialogue.