Made with 🧠 and 🫀 by Youssef Bouksim

Back to library
⏳

Labor Illusion

We value results more when we can see the work being done to produce them. The same output feels more valuable after watching it be created than after receiving it instantly. This has nothing to do with the quality of the result — only with our perception of the effort behind it.

5 min readLoading States · Perceived Value · Trust

In 2011, researchers Ryan Buell and Michael Norton ran an experiment with a travel search site. They showed users search results in two conditions: instantly, and after a delay during which a progress bar showed the system “searching 100+ sites.” Users in the second condition rated the results as better and were more satisfied — even though the results were identical. The delay wasn't a technical limitation. It was a design choice. And it made the output feel more thorough, more trustworthy, and more valuable.

Buell and Norton called this the labor illusion: the tendency to value service more when we can see the effort being put into it. The visibility of effort creates the perception of quality, independently of the actual quality.

This finding has direct consequences for loading states, AI response generation, search interfaces, and any part of a product where processing time is involved. Instant delivery, paradoxically, can make results feel less trustworthy than the same results delivered after visible processing.

✦ Three things to know
✓
Perceived effort predicts perceived quality.A search that shows “checking 847 flights” feels more thorough than one that simply returns results — even if both searched the same inventory. The displayed work creates an inference about quality that is automatic and largely unconscious.
✓
Showing steps is more effective than a progress bar.A sequence of meaningful steps — “analysing your preferences,” “comparing 1,247 options” — conveys both duration and work. Specificity is what creates the value perception, not the time itself.
✓
The illusion works only when the work shown is plausible. Manufactured effort signals deception, not quality. The line between design and dark pattern is whether displayed effort bears some relationship to actual processing.
“Showing your work isn't just pedagogy. It's a signal of diligence that shapes how the output is evaluated — before the output even arrives.”
— Ryan Buell & Michael Norton, 2011

The same result — felt differently depending on how it arrives

The simplest way to see the labor illusion is to receive the same output two ways: instantly, and after watching it being produced. The output doesn't change. What changes is how you feel about it. Tap both buttons below and notice which result you trust more — and why, given that they're identical.

Before — Instant delivery
9:41
FlightsInstant
Query
Find me a flight from Casablanca to Paris, cheapest option, next weekend.

No evidence of work done. The answer feels shallow even though it is the same.

After — Visible effort
9:41
FlightsVisible effort
Query
Find me a flight from Casablanca to Paris, cheapest option, next weekend.

Fourteen airlines checked, 387 combinations compared. The same answer now feels earned.

The result is the same flight at the same price. But the working version has earned it — the user watched the system check 14 airlines, compare 387 combinations, and filter by fastest connection. This is why Kayak, Skyscanner, and every major travel search engine shows searching animations even when results could be returned faster. It's a design choice that produces the labor illusion — and with it, a stronger perception of thoughtfulness and quality in the output.

AI tools — streaming as labor illusion

The most prominent current application of the labor illusion is in AI products. When Claude, ChatGPT, or any LLM streams its response word by word, the technical reason is that tokens are generated sequentially. But the design effect is a labor illusion: the user watches the thinking happen. The response that builds itself in front of you feels more considered than an identical response that appeared all at once.

The two response displays below contain the same analysis. Tap both and notice which feels more trustworthy.

Before — Instant response
9:41
AIInstant
Prompt
“Analyse the main risks in this business plan.”

Text appears in full. Reads as a machine answer, not a considered one.

After — Streamed response
9:41
AIStreamed
Prompt
“Analyse the main risks in this business plan.”

Tokens arrive sequentially. The same output now feels like thinking made visible.

Both analyses are identical. Users consistently rate streamed AI responses as more thoughtful, more thorough, and more trustworthy than the same text displayed instantly — even when they know the model generated the full response before streaming began. The emotional evaluation of the output is shaped by the viewing experience, and that shaping is remarkably resistant to rational correction.

When to use it — and when it backfires

The labor illusion is most effective when three conditions are met: the work shown is plausibly related to the output, the delay is within the range of expected processing time for the task, and the user has no strong prior reason to believe the progress display is fake. Break any of these and the effect reverses — instead of increasing trust, the visible “work” signals deception.

The classic backfire case is a loading animation that plays for the same duration regardless of the input. Users who notice this — and frequent users notice it quickly — lose not only the labor illusion but their baseline trust in the product's honesty.

✓ Apply it like this
→Show specific steps, not generic spinners — "checking 847 flights across 14 airlines" produces stronger labor illusion than "loading."
→Vary the loading behaviour with the input — a search that always takes exactly 2.3 seconds is obviously fixed. Variable timing maintains plausibility.
→Stream AI and computation outputs where possible — displaying tokens as they're generated is genuine, and produces the labor illusion as a side effect of honest real-time display.
→Use the wait time productively — show relevant tips, context, or previews that make the delay feel purposeful rather than just a wait.
✗ Common mistakes
→Fixed-duration fake loading — if the same animation plays for identical time regardless of complexity, users will notice and interpret it as deception.
→Progress steps that don't match the output — "analysing your writing style" for a tool that doesn't customise by user breaks plausibility and produces distrust.
→Artificial delays on frustrating tasks — adding delays to form submissions or error recovery compounds frustration rather than adding perceived value.
→Overclaiming the work — "searching 10 million sources" when the actual index is 10,000 is a verifiable lie. Users who check lose all trust beyond loading screens.

Buell, R. W., & Norton, M. I. (2011). The labor illusion: How operational transparency increases perceived value. Management Science, 57(9), 1564–1579. · Norton, M. I., Mochon, D., & Ariely, D. (2012). The IKEA effect. Journal of Consumer Psychology, 22(3), 453–460. · Buell, R. W. (2019). Operational transparency. Harvard Business Review, 97(2), 102–113.