Deciding

Loss aversion

Giving something up feels heavier than gaining the same thing felt good.

Losses and gains of equal size are not weighed equally. In prospect theory the value function is steeper below the reference point than above it, so a loss of a given size hurts more than an equivalent gain pleases. What counts as a loss depends entirely on where the reference point is set, which is why the same outcome can be framed either way.

The same amount, weighed twice

A gain and a loss of identical objective size do not carry identical psychological weight. In prospect theory the loss side sits lower, which is why removing a feature provokes more reaction than adding an equivalent one. The ratio here is illustrative, not a measured constant: published estimates vary widely and some researchers dispute that a stable ratio exists at all.

How it shows up in software

loss aversion: Losing something feels worse than gaining the same thing feels good. is why cancellation flows list what the user is about to lose and why free trials convert better once a user has set up real data inside them. It also explains streak mechanics: after day 14, the streak is a possession, and breaking it registers as a loss rather than a missed gain. The same logic makes users resist migrations that would leave old work behind.

Using it well

During a trial, help users create something they would miss. A populated workspace is a reference point; an empty one is not.
In a destructive-action confirm, name the specific thing being lost ('12 saved reports') instead of asking a generic 'are you sure'.
When you remove a feature, offer a real replacement path in the same message, so the change reads as a swap rather than a subtraction.
Test gain framing and loss framing on the same message. Loss framing is not reliably stronger, and it often reads as threatening in a work tool.

Where it turns manipulative

Confirmshaming is loss aversion turned into guilt. A decline button reading 'No thanks, I like wasting money' is a named dark pattern: An interface built to get an outcome the user would reject if it were stated plainly. and is treated as an unfair practice under the EU Digital Services Act.
Manufacturing a loss that does not exist, such as a countdown that resets on reload or a 'you will lose your data' warning when the data is retained for 30 days anyway.
Streaks built on daily obligation can push anxious usage. If losing the streak hurts more than using the product helps, the mechanic is working against the user.

Where you have seen it

Duolingo
The streak counter is shown at the top of the home screen, and a streak freeze item can be bought or earned to protect it.
Dropbox
Downgrade flows show which files would exceed the free plan's storage limit before the downgrade is confirmed.
Figma
Deleting a file moves it to a trash area with a stated recovery window rather than removing it outright.

What the research says

Kahneman and Tversky, 1979Well evidenced
Choices over gambles showed systematic risk aversion for gains and risk seeking for losses, which the authors modelled with a value function that is kinked at the reference point.
Kahneman, Knetsch and Thaler, 1990Mixed evidence
In repeated market experiments, people given a mug demanded roughly twice what buyers were willing to pay for it, and trading volume stayed far below what standard theory predicts.
Gal and Rucker, 2018Contested
A review argued that much of the evidence for losses looming larger is better explained by inertia, status quo preference, or the way options were presented, and that the 2:1 loss-to-gain ratio is not a stable constant.
The direction of the effect holds up in many settings. The size does not, and it is not a fixed multiplier you can plug into a model. Treat loss framing as a hypothesis to test, not a lever with a known gain.

Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.