How Many Meanings Live Inside One Word: "Risk"── When the reviewer's "risk" and the maker's "risk" point at different things

Part 16. This installment works as a kind of blueprint for the series: it shows how the single word "risk" splits apart in real work. People use the same word as the name for entirely different things. I open up, one by one, the cracks I keep hitting every time I check a piece of promotional material.

01The same "this is risky" pointed at different things

In a meeting about a piece of material (= the document used to explain a drug's effects and proper use to doctors), we got stuck over one phrase. The description of the drug's effect had a single line in it. I thought keeping it would invite misreading, so I said we should cut it. "This wording is too risky." The maker shook his head at once. "No — cutting it is riskier. A doctor who reads it will end up underrating the effect."

The meeting ran in parallel lines for about ten minutes, and that day it went back for revision. On the way home, it stayed lodged in me. We had both said "this is risky." The same words. Yet the conclusions came out exactly opposite. One of us meant "keeping it is dangerous," the other "cutting it is dangerous." Neither was lying. At my desk, I wrote out in a notebook what each of us had meant by that one word.

The "risk" I meant was the chance of being misread and leading to inappropriate prescribing. The "risk" he meant was the chance that correct information fails to reach the doctor and the patient loses out. The contents we pointed at were different from the start. We were using the same three letters as names for different things. Until I noticed that, we kept talking past each other and only raised our voices.

I learned later that the maker wanted that line kept for a reason. On an earlier, similar material, he had cut the explanation, and afterward doctors kept calling in to ask "what do we do in this case?" — and he was left scrambling to respond. Whether to include the line carried, for him, the weight of a bitter past experience. Behind the single sentence I had simply read as "saying too much," he had his own history.

One researcher who studied why an exchange of words works at all is Herbert Clark (= a psychologist who examined how conversation operates). He proposed the idea of common ground (= the body of knowledge speaker and listener both assume "we already share this"). A conversation rides not on the words themselves but on the shared assumptions beneath them. When the assumptions line up, even short words get through. When they don't, the same word points at different things inside two heads.

What was missing in that meeting was not vocabulary. We both knew the word "risk." What was missing was the groundwork of agreeing on what the word pointed to inside this particular material right now. In the margin of my notebook I wrote, small: "When you clash over the same word, first doubt the dictionary — the other person's dictionary."

02Five contents inside one word

While going back over the next material, I tried, as an experiment, circling every "risk" in the text with a pencil. Within a single document it came up again and again. And as I circled, I noticed that the same characters pointed at things that shifted slightly each time. It looked like one word, but the contents were not one thing.

When I sorted them out, it came to roughly five.

Reread that sent-back meeting through these five and the scene changes. My "this is risky" was close to ④. I meant the chance that misreading spreads and a bad outcome follows. But his "this is risky" was not the same ④. It was closer to ①: the severity and likelihood, bundled together, of patients being harmed because the information fails to arrive. I had pulled the same three letters from the ④ drawer, he from the ① drawer. The harm we pointed at, and the way of counting it, were both different. That is why we did not mesh.

It is not that the word has no fixed rule. There is an international agreement, ISO 31000 (= a global standard on how organizations should handle risk), where risk is defined in one line as "the effect of uncertainty on objectives." A well-honed definition, I think. But the "risk" people actually say on the floor refuses to fit neatly into that line. One person packs probability into the three letters, another dread, another the source of danger itself.

A definition being settled as one thing, and the meaning being lined up as one thing inside people's heads, are two different matters. Precisely because the standard can be written in one line, we fool ourselves into thinking "we already share it." The root of the confusion is not too few words but too much packed into one word. When I check a material now, the moment I find the word "risk" I stop once and ask: which of the five is this? Circle it, number it. That alone clears away half the misunderstandings before the meeting even starts.

03Same probability does not mean same dread

In a corner of one material, a very low-probability side effect was written in a single line. In numbers, maybe one in several thousand. I was about to let it pass as is, but the maker, facing that one sentence, plainly tensed up. "Written this way, the doctor reading it will brace far more than necessary." The probability is near zero. Even so, for him the smallness of the number did not translate into reassurance. That was when I realized: the same sentence, I was reading by probability and he was reading by dread.

One researcher captured this gap as numbers. Paul Slovic (= an American researcher who measured the perception of risk through psychology) asked large numbers of people how big they felt various dangers to be, and organized the answers. What moved the feeling was not only how likely an accident was. Two axes had a strong effect. One was dread (= the sense of being uncontrollable, inescapable, harming many at once). The other was the unknown (= the sense that the mechanism is unclear, invisible, with effects that appear later). The higher these two run, the larger people rate the danger at the same probability. Many feel nuclear power is "scarier" than car accidents. But the reason is not the number of deaths — it is these two axes. So Slovic read it.

That one line about a side effect landed exactly there. The chance of it happening is low. But the feeling of "I can't protect myself from it" and "by the time you notice, it has already progressed" pushes up both the unknown and dread. The person tensed up not from a miscalculation but because people are simply built to feel that way.

The third, like-or-dislike, has its own name: the affect heuristic (= a shortcut where the first impression of liking or disliking runs ahead of and steers the judgment of dangerous-or-safe). When a person feels a method is "agreeable," they rate its danger low and its benefit high. The reverse happens too. So the danger estimate is not the result of cold calculation; the reasoning catches up afterward to the first feeling. I think the person's dislike of that sentence was, half of it, this first impression.

What I learned here is this. "If you write the probability correctly, it gets through" was a belief of the maker — and at the same time a belief of mine, the reviewer. The same word "risk," I read as a fraction and he read as a stomach-tightening sensation. Neither is wrong. What the one word pointed at was split in two from the start. Push only the correctness of the numbers without noticing that, and the other person either goes quiet and backs off or gives way somewhere else. The first source of the gap is right here.

04What you count loss from is out of line

With the same revision proposal between us, the two sides of the desk can reach opposite conclusions. I marked up one material in red. I toned a phrase down a notch and added a line of conditions. In my head this was loss-prevention work. Sent out unfixed, a doctor who believes the overstated explanation might use it on a patient it doesn't suit. That loss can't be taken back. So I fix it.

But the maker, looking at the same red, was counting a different loss. Toned down, the information he truly wanted to deliver thins out. Adding the condition increases the labor of reprinting and replacement. A replacement right before sign-off especially pushes the whole reproof-and-print schedule backward, and the day the MRs can hand the material out on the floor slips to the next distribution cycle. For them, that was exactly the loss to avoid. Both of us are trying to avoid loss. Yet the answers face opposite ways. I say "fix it," he says "don't." It is not a matter of good or bad — what we count loss from was different from the start.

The pair who put this gap into words are Daniel Kahneman and Amos Tversky (= two psychologists who showed biases in judgment through experiments). Their prospect theory (= an account of how people feel loss and gain differently) has three tools that bite here.

• Loss aversion (= for the same size, the pain of losing is felt about twice as strongly as the joy of gaining). So both of us reach to avoid loss rather than to grab a gain. The direction of playing defense is the same, but what we are defending differs.
• Framing (= the same fact changes how people choose depending on whether it's stated as a gain or a loss). "90 of 100 used it without trouble" and "10 of 100 it didn't suit" have the same content, but the dread received is utterly different. My one red-penned line, too, drew a different reaction depending on which phrasing I chose.
• Reference point (= the starting line from which loss or gain is counted; above it looks like gain, below it like loss). My starting line was "a correct material is out in circulation"; his was "the finished draft in hand right now." Shift the starting line and the same edit looks, to one, like avoiding loss, and to the other, like having something taken from what they hold.

The third, the reference point, runs deepest. The same edit of one sentence changes shape into either a gain or a loss depending only on where you place the starting line. The very operation I think of as "fix here and we move back to the safe side," he feels as "something is being subtracted from what is already finished." The instant it feels like being subtracted from, loss aversion kicks in and resistance hardens. What I want to show is a step toward safety; what he feels is the pain of a possession shrinking. The same stroke of the red pen pointed at different things on the two sides of the desk.

So when I mark in red, I began to add a word. "This isn't to cut something — it's an addition so we don't carry a different loss later." Lining up the starting line, only that much. But unless the starting lines match, no matter how carefully I fix, it can only look to the other person like their own work being shaved away. The second gap in the word "risk" lurked not in probability or dread but right where you set the starting point for counting loss.

05The failure that comes to mind easily looks bigger

Last month, I found the line "side effects are almost nonexistent" in a material and marked it heavily in red on the spot. It declared a degree with no data behind it. The back-and-forth with the maker dragged on, and in the end it went back amid an awkward mood. Ever since, whenever I see words of degree like "almost" or "rarely," my shoulders tighten. When that kind of expression shows up, there is a self in me bracing "this is dangerous" before I've even read the content.

That bracing is usually overdone. Count it calmly, and most words of degree passed without trouble. Yet in my head that one incident was blown up large, building the feel that "this kind of expression often causes trouble." Not how often it actually happened, but how easily it came to mind was pushing up my danger estimate.

The psychologists who put this into words are Amos Tversky and Daniel Kahneman. They showed the availability heuristic (= the mental habit of feeling that events easier to recall are more likely than they really are). Just as you feel afraid of flying right after seeing news of a plane crash, one strong memory alone makes people rate that kind of event as more probable. The vividness of memory gets swapped in for the sense of likelihood.

The trouble is that the maker and I remember different failures. I vividly remember "the one time I clashed over declaring a degree." Meanwhile the maker may still be carrying "the time a roundabout way of writing an effect got him sent back over and over." Facing the same material, the two of us brace against different mines. While I sharpen my nerves over words of degree, the maker worries about the fineness of phrasing. The danger estimates fall out of line not from a difference in ability but from a difference in the content of memory.

So I make it a rule to doubt my own bracing once. When I feel "this is dangerous," I separate whether it's a judgment from the data in front of me or an echo of last month's incident. I go back over my own check records and confirm what fraction of the time that expression actually caused trouble in the past. I return the judgment from impression to counted fact. Do that, and I find plenty of spots where I had been overtensing.

06Write it as a %, or as so-many-in-100

The same side effect in the same promotional material can be written two ways. "It happens in 5%" or "5 in 100 people." As numbers they are exactly the same. But how the reader takes it clearly changes. What I think over and over while I read these through is that the % notation looks clever but doesn't get through. Even when it says "5%," the reader skims past it without forming a picture of how many patients that is in their own clinic. This gets harder as the rate gets smaller. With "0.1%," turning it back into people in your head suddenly becomes difficult. And when two rates stack up — "it happens in 5%, and of those 20% are severe" — even an experienced reader can misread it. It isn't laziness. We are simply not built to turn a vague number back into real, living people.

The psychologist Gerd Gigerenzer studied this gap in reception. He showed that people understand natural frequencies (= writing it as an actual head count, like "5 in 100 people") more correctly than probabilities (%). A % is the result of a division, and it takes another step in the head to turn it back into a head count. Somewhere in that conversion, people get it wrong. With "5 in 100 people," that step isn't needed. Just changing how it's written delivers the meaning without making anyone calculate.

One more thing I sense as a fuse for sent-back revisions is how to write a danger whose numbers are unknown. People dislike a danger of unknown probability more strongly than one of known probability. Daniel Ellsberg (= a decision researcher who studied through experiments how people treat the uncertain) showed this experimentally. People prefer to avoid a bet of unknown odds over one of known odds. This is called ambiguity aversion. People prefer drawing a lottery from an urn whose contents they can clearly see over one whose proportions are unknown. Even at the same chance of winning, the not-knowing itself is disliked.

So when a material has only words of degree and no number, I always stop. "Rarely" and "occasionally" make different people picture wildly different head counts. One imagines 1 in 1,000, another 10 in 100. That gap becomes, later, the complaint "this isn't what I was told." What I ask of the maker is to put known numbers in head-count form, and when they're unknown, write honestly that they're unknown. Hidden ambiguity is what's disliked most. A clear note that "there isn't enough data here yet" lets the reader take it in more calmly.

That said, what I loosen is only the strain in the expression. A flat assertion of "almost none" with no data behind it, or a word of degree that different readers split on, doesn't get a pass just because it's a matter of phrasing. Let that through and a material that overstates by mistake goes out into the world. I lower the strain, but I don't lower my eye on thinly grounded assertions and vague words of degree. That one line I do not move.

07Building the ground where talk gets through even across positions

Looking at the same material, the danger the maker and I see as large is out of line. Briefings keep making me notice it. I worry first about "could this wording be taken to mean the drug works too strongly?" The maker worries first about "if we cut here, won't important information that should reach the patient drop out?" Both of us seriously have the patient in mind. Yet the content of what we worry about is different.

Why can't we stand in the same place? The cultural anthropologist Mary Douglas (= a researcher who read out, from a group's culture, how people sort things into "clean / dangerous") wrote that people see a danger that fits their own group's way of seeing as larger than it is. The legal scholar Dan Kahan remeasured this in present-day terms and named it cultural cognition (= the mental habit of feeling a danger larger when it fits the values of one's own circle). Shown the same fact, people take in strongly the story that troubles their own position and take lightly the story that benefits it. How the data is read bends with position.

Place this in the world of materials and it becomes clear. As the reviewer, I stand where "if I get blamed after it's gone out, that's trouble." So I see large the danger of things coming across as overstated. The maker stands where "if information the doctor or patient needs doesn't arrive, that's trouble." So they see large the danger of information running short. Neither of us is being lazy. The place we stand is choosing which danger we see.

What's troubling is that this gap hides inside the one word "risk." When I say "this is too risky," I'm pointing at the danger of overstatement. When the maker comes back with "no — cutting it is riskier," he's pointing at the danger of information running short. With the same word, each pointing at something different, we keep pushing. We aren't even raising our voices, yet the talk doesn't mesh.

What works here is the idea of Herbert Clark, the psychologist who studied the exchange of words. He said a conversation holds together because the two speakers share "what are we talking about right now," and he called that shared footing common ground (= a state where the two can confirm to each other they're pointing at the same thing). When it slips, you confirm again in words, each time. Clark observed that people are constantly re-confirming inside a conversation. What our briefings lacked was exactly that one extra step.

So at the start of a briefing I began to say this: "What I'm about to talk about is the danger of looking too strong. The danger of information running short, let's take that separately, afterward." Just this much, and the air changes. The other person stops bracing, and I find it easier to put my own blind spot on the table. Cultural cognition can't be erased. The bias in the danger a position shows you follows you as long as you're human. If it can't be erased, first say aloud and match "which meaning of risk are we talking about now." That was, on the floor, the most effective single move for letting two people in different positions look at the same target.

What this series has looked at all connects here. In the earlier installments, I lined up real cases of materials going astray and saw how a small slip of words becomes the doorway to a large misunderstanding. In Part 13 I wrote that the reviewer and the maker are not two opponents but companions standing side by side toward the same risk. In Part 14 I took up the power to notice a dangerous sign early. And now this time: when the two talk over a danger they've noticed, first match "which meaning of risk." Stand side by side, notice, match the words. Only when these three line up does the talk get through even across positions. Aren't we pointing at different things with the same word? I always start from confirming that first.