The reason a search box changes what people want
Autocomplete offers a finished question before the user has finished asking one. The suggestion is usually accepted, and it was not the original query.
Filed by The Archivist 2 min read
Intuition test — answer before you read on
Users frequently accept an autocomplete suggestion instead of finishing their own query. What is the consequence?
Correct answer: B
Suggestions are drawn from frequent past queries, so accepting one both saves effort and feeds the frequency that generated it. The user never sees what their own phrasing would have returned, so the substitution has no visible cost.
A user types four characters. Six completed queries appear. One of them is close enough to what was intended, and it is selected in preference to finishing the original thought. The search that runs is the platform’s phrasing of the question, not the user’s.
What everyone sees
Autocomplete is experienced as convenience and speed, and it is both. Users describe the suggestions as reading their mind, which is a reasonable description of a system trained on the queries of people like them. The substitution is not noticed because the suggestion resembles the intention closely enough to feel identical.
What is actually happening
Query suggestion narrows the space of asking. Suggestions are drawn from popular past queries, so they systematically favour questions that have already been asked often, and selecting one is faster than typing an unusual phrasing. Research on query reformulation and suggestion bias, including work by Baeza-Yates on search behaviour and by Introna and Nissenbaum on the politics of search, describes the resulting feedback loop: common queries are proposed, proposal makes them commoner, and rare formulations become progressively harder to reach.
Why it stays hidden
The narrowing hides because the accepted suggestion satisfies the user. Whatever the original phrasing would have returned is never seen, so there is no comparison to be disappointed by. The interface also presents suggestions as help rather than as a shortlist, and help is not the kind of thing users audit for what it left out.
Suggestion shapes the question. What is easy to ask becomes what gets asked, and the alternative results are never observed.
Suggestion shapes the question. What is easy to ask becomes what gets asked, and the alternative results are never observed.
Collect this card
Suggestion shapes the question. What is easy to ask becomes what gets asked, and the alternative results are never observed.
0 / 10,000 collected
Sources & further reading 3
- Introna & Nissenbaum, "Shaping the Web: Why the Politics of Search Engines Matters", The Information Society, 2000
- Baeza-Yates, "Bias on the Web", Communications of the ACM, 2018
- Sullivan, "How Google Autocomplete Works in Search", Google, 2018
Cross-references
Related files
Filed near this one in the index.
-
Hidden Mechanics Entry #0008
The hole in your pen cap is a lung
That small hole in the top of a ballpoint cap is not for airflow to the ink, or for pressure, or for manufacturing. It is there because people chew pens and occasionally swallow the cap.
Adept -
No visual on filePlatform Mechanics Entry #0287
The reason a platform’s rules change after you build on them
The rules were generous when they needed builders. Once the builders are invested, the rules can be revised, because the cost of leaving exceeds the cost of complying.
AdeptThe hidden part #0287Generous terms attract investment. Investment creates dependency. Dependency permits revision. The sequence is not accidental.
Platforms Open file -
No visual on filePlatform Mechanics Entry #0284
Why network effects make second place worthless
The value of the first network rises with each user. The value of the second rises too, but the gap between them widens rather than closes.
AdeptThe hidden part #0284The first network large enough owns the people. The second can copy every feature except the one that matters.
Platforms Open file -
No visual on fileHabit Loops Entry #0302
The reason a habit survives the loss of its reward
The reward that started the habit can vanish entirely, and the habit continues. What was once a means to an end has become the end itself.
AdeptThe hidden part #0302Repetition outlives reinforcement. The habit that was built by a reward no longer needs it, and no longer checks.
Habit Open file
Annotations are reserved for archive members.
Sign in to annotate