The Nudge That Backfired On Its Own Designer

I've been setting behavioural traps for myself. Running shoes by the front door. Sticky note on the fridge. Phone banished from the bedroom at 9pm. Water glass next to the coffee machine...placed there deliberately, strategically, by a psychologist who knows exactly how habit formation works.

And then I've been walking past them. Moving the phone. Drinking the coffee, ignoring the water, leaving the shoes where they sit. Feeling guilty. I know what the research says. I designed the nudges to help change my behaviour. And they failed anyway.

Here's what's interesting about that.

Why nudges became a big deal, and what got skipped over

A nudge (the term comes from Thaler and Sunstein's 2008 book) is any small change to your environment that makes the better choice easier — without forcing anything or changing the economics. Fruit at eye level in the cafeteria. Opt-out rather than opt-in pension enrolment. A painted fly on a urinal at Schiphol Airport that reduced spillage by 80%. (Yes, really.)

Governments loved them. Cost-effective, politically painless, no legislation required. The UK's Behavioural Insights Team ran hundreds of real-world applications. Dozens of countries followed.

For a while, the story was almost entirely triumphant.

The assumption nobody examined closely enough

The foundational promise of nudge design is this: they help people who have behavioural blind spots, and they're essentially harmless for everyone else. You can always opt out. At worst, the nudge just doesn't apply to you.

A decade of research has been dismantling that promise.

Linda Thunström at the University of Wyoming ran the numbers on calorie menu labelling (the classic salience nudge that's now mandatory across the US). For people with high eating self-control, the labels worked fine. Useful, low-cost, mild improvement in food choices.

For people with low eating self-control (i.e the exact people the nudge was designed for), the calorie information acted as what Thunström called an "emotional tax." Anxiety. Guilt. Worse affect. And less behaviour change than in the group who didn't need it.

The nudge worked on average. It helped the people already eating well. It distressed the people who weren't and moved their behaviour less. That's not a rounding error. That's the headline finding hiding a story that runs almost exactly backwards.

Her spending research using Scott Rick's spendthrift framework is even stranger. A prompt reminding people to think about opportunity cost before a purchase e.g. "what else could you buy with this?", reduced spending. But only in people who already felt too much anxiety about money. The actual overspenders were essentially unaffected. Then she flipped the nudge, redesigning it to increase spending in people who were underspending. Same result. Both nudges pointed in opposite directions.

The average effect was positive. The effect on the people who needed it most was zero, or negative.

What happened with my running shoes

What I experienced has a name: psychological reactance. When a person perceives a threat to their behavioural freedom, even a mild one, even self-imposed, even technically optional, our brain moves to restore that freedom. Sometimes by ignoring the constraint or doing the exact opposite.

Jack Brehm first described this in 1966. Sixty years of subsequent research has confirmed it applies to nudges specifically. A 2023 paper on food choice found that when participants perceived a nudge as intruding on their autonomy, a backfire effect could emerge almost immediately. The same nudge that works reliably on low-reactance people produces the opposite behaviour in high-reactance ones.

Some people are dispositionally more reactive to feeling steered than others. And because high-reactance people are often more aware of being influenced, covert nudges (the subtle kind designers usually prefer) can actually make it worse.

I am, I suspect, fairly high in trait reactance. My own nudges feel, to some part of my brain, like being told what to do. That part has a very reliable response.

It says: it’s a no from me...

The fix that actually works (at least for me)

The most interesting response to this in the current literature is something called nudge+, aka. transparent nudging. Instead of hiding the mechanism, you name it. You tell people what the nudge is and why it's there.

Early results are genuinely good. Transparent nudges where participants are explicitly told "we've set this default to make the sustainable option easier" produce equal or greater behaviour change than covert ones, with significantly less reactance. The transparency doesn't undermine the effect. Often it enhances it, because the person isn't being steered without their knowledge. They're knowingly choosing to participate in a system they understand.

That distinction is load-bearing for the brain.

For my running shoes: the version that actually works isn't the shoes silently waiting. It's telling myself explicitly, the night before "I'm putting these here because morning-me will look for reasons not to go, and I'm choosing now to make it slightly harder to opt out." Same mechanism but my autonomy is preserved and my reactance, so far, lower.

Three things worth taking seriously

If you're designing behaviour change for a team, an organisation, or just yourself here's what to take away:

• Average effects hide distributional stories. A nudge that improves outcomes overall can actively harm the subgroup it was designed for. Before rolling something out, ask: who might this hurt? Disaggregating by subgroup tells you what the headline number doesn't.

• Autonomy isn't incidental. It's the whole mechanism. Nudges work because they preserve choice. They backfire because they violate the felt sense of autonomy, even when the formal choice hasn't changed. Covert nudges that get detected often produce more reactance than transparent ones would have from the start.

• Self-nudging requires self-knowledge. If you're high in reactance, i.e. if you bristle at feeling managed, (even by yourself!) standard nudge design applied to your own behaviour may actively undermine what you're trying to build. The answer isn't cleverer nudge design. It's a different relationship with the nudge: explicit, chosen and visible.

The "nudges are harmless at worst" assumption needs to retire. They're not always harmless. Sometimes they help the wrong people. Sometimes they push back hardest on the people who most need the push.

The next generation of behaviour change design needs to be honest about that and more interested in the person in front of it than the average effect across everyone.

Over to you! What nudges are currently running in your organisation for spending, eating, coming to the office, working, communicating and do you actually know what they're doing to the people they weren't designed for?