You pick a trend you like. You go looking for evidence. You find ten great examples. Congratulations: you've proven that you like it.

That's the trap at the start of almost every trend project, and a lot of design research too. The algorithm shows you more of what you already engage with, and your brain holds on to whatever confirms what you already believed. So before learning how to spot trends, it helps to learn how your own head gets in the way.

What a cognitive bias is

A cognitive bias is a mental shortcut the brain uses to decide quickly, and that produces the same kind of error over and over.

You've felt one. You walk out of a shark movie and the ocean suddenly feels dangerous. The odds of a shark attack didn't change. What changed is how easily the image comes to mind.

Amos Tversky and Daniel Kahneman showed in "Judgment under Uncertainty: Heuristics and Biases" (1974) that these errors aren't random and aren't just carelessness. They're predictable, and they show up in experts as much as in beginners. That's exactly why they're worth studying.

One misunderstanding to clear up: a bias isn't the same as an opinion. You can defend an opinion. A bias works without you noticing, which is why thinking harder doesn't fix it. Method does.

The four that do the most damage

Confirmation bias

You search for, notice and remember what confirms what you already believe. Raymond Nickerson called it "a ubiquitous phenomenon in many guises" (1998): it turns up in science, in medicine, in juries. In trend work it looks very ordinary. You choose a trend, then the ten signals you collect are, surprise, the ten that support it.

Availability

You judge how likely or how big something is by how easily an example comes to mind. Whatever showed up in your feed this week feels bigger than it is.

Anchoring

The first number or idea you receive sets your range. If the first report you read says a market is worth forty billion dollars, you'll read everything afterward against that number, even if the report was wrong or had a reason to inflate it.

Novelty bias

You prefer the new because it's new. This one isn't in the classic Tversky and Kahneman list, but it's the designer's professional bias, and it's the one that disguises itself best as "good taste."

All four push in the same direction: toward declaring a trend out of something you already liked.

A few more worth knowing

Status quo bias. William Samuelson and Richard Zeckhauser (1988) showed that people disproportionately stick with the current option, even when switching would be better. In research, it shows up as dismissing a signal because "people will never change how they do that." They often do.

Overconfidence at low skill. David Dunning and Justin Kruger (1999) found that people who performed worst on tests of logic, grammar and humor also overestimated their performance the most, partly because the skills needed to do well are the same ones needed to judge how well you did. In trend work, the less you know about a field, the more obvious its future looks.

Biases built into the models themselves. Everett Rogers' Diffusion of Innovations is the classic model of how ideas spread: innovators, early adopters, early majority, late majority and laggards. It's useful because it tells you where to look for early signals. But Rogers himself names problems in his own field:

• Pro-innovation bias: assuming every innovation should be adopted, and quickly.
• Individual-blame bias: blaming the person instead of the system when they don't adopt something.
• The recall problem: people remember badly when they adopted something, so the data about timing is shaky.

The word "laggard" is individual-blame bias in a single word. Before you use it, ask whether that person actually rejected the thing, or simply couldn't afford it.

Bias in the sources, not just in you

Your sources have their own pull. Two habits help.

Read platform numbers for what they measure. Social media reach figures, like the ones compiled in DataReportal's country reports, come from what platforms declare as reachable ad audiences, not from counted people. Duplicate accounts, business accounts, old accounts and misattributed locations all get included. That's how you can end up with a platform "reaching" more adults than a country actually has. The number is useful. It just doesn't say what it seems to say.

Say who publishes and how they make money. Take "expectation economy," the idea that customer expectations only rise because every good experience in one industry becomes the new minimum everywhere else. If a delivery app gets food to you in fifteen minutes, your bank starts to feel slow. It's a sharp observation, and it comes from TrendWatching, a consumer trend agency that sells reports and subscriptions to brands. So cite it as a live industry source, not as a verified finding. A market forecast made by someone who lives off that market reads differently, and your readers deserve to know.

You can't switch a bias off. You can design around it.

There's no setting that turns a bias off. What you can do is build a process where your biases have to pass through a filter. These four rules do most of the work:

1. Get out of your feed. At least two out of every five signals should come from outside the internet: the street, a shop, public transport, a classroom. If they all came from one screen, you documented an algorithm.
2. Date everything. Without a date, you can't verify it. The date is what separates "I saw it coming" from "I remembered it afterward."
3. Look for the counterexample. For every trend you propose, actively hunt for the case that contradicts it. If you didn't find one, you didn't look.
4. Write down your bet before you research. Note what you expect to find. If at the end you found exactly that, be suspicious of yourself.

The same rules apply to design research in general. Interview notes without dates, insights that happen to match the stakeholder's favorite idea, a persona built from the three users who were easiest to reach: same biases, different room.

Try it

Before your next research round, write one sentence: "I expect to find that..." Seal it away. Collect your evidence with dates and at least some of it from offline. Then compare. If what you found matches your sentence perfectly, go looking for the counterexample you skipped.

Further reading

• Tversky, A. and Kahneman, D. "Judgment under Uncertainty: Heuristics and Biases." Science 185(4157), 1974.
• Nickerson, R. S. "Confirmation Bias: A Ubiquitous Phenomenon in Many Guises." Review of General Psychology 2(2), 1998.
• Samuelson, W. and Zeckhauser, R. "Status Quo Bias in Decision Making." Journal of Risk and Uncertainty 1, 1988.
• Kahneman, D. Thinking, Fast and Slow. Farrar, Straus and Giroux, 2011.
• Rogers, E. M. Diffusion of Innovations. Free Press, 5th ed., 2003.