[Swizec]

You raise an interesting question. But ultimately the wrong kind of question. We shouldn't be asking ourselves how early can humans effectively search the internet. We should be researching how soon can humans make a search engine that is natural enough a small kid can use.

Now I don't have any kids or access to anything of the sorts, but isn't 6 exactly the age kids start asking questions? Questions that are sometimes hard to answer. This indicates that kids are in fact avid users of search, they just aren't very good at translating their query into something google can usefully understand.

Have you ever tried searching for something you didn't know how to exactly specify? My recent example is seeing a cool car in London. Now that I know what it's called I can just search for "morgan three wheeler". But the first time 'round I started from "three wheel oldtimer" and probably a few others like that, then looking through a lot of wikipedia etc.

Until we can make search engines behave like asking a human "Hey, that car looked cool, what is it?" kids won't be avid searchers.

[frossie]

isn't 6 exactly the age kids start asking questions?

Yes but a 6 year old still believes their parents :-)

Seriously, the issue is that the questions young kids ask need a lot of context to interpret - they are not the kind of thing you can type at a search engine. For a (real) example a kindergartener watching a cartoon might ask "why are they walking normal"? So the parent has to figure out that (a) the cartoon characters are supposedly on Mars (b) the kid thinks Mars is like the moon (c) the kid has seen footage of real astronauts on the moon (d) ergo the kid had an expectation for the characters to walk "funny" on Mars.

The kid is asking a good question, but is not able to form the googlable question "What is the gravity on Mars"? It needs an adult (and frequently specifically a familiar caregiver) to mediate the question it is asking to the question that can be answered.

[scott_s]

Before we do any of that, we have to determine if children are already effective with current search engines. It's possible you want to build something no one needs. The study linked here did not establish that.

Sometimes, if my understanding of what I want is fuzzy, or I just don't want to expend the mental effort to synthesize good search terms, I'll just pass a naive question to Google. It often helps me down the path to determining better keywords, and occasionally it finds what I'm looking for.

In other words: what you want might already exist. We don't know until we study it.

[metageek]

>isn't 6 exactly the age kids start asking questions?

Uh, no. My kids were asking good questions by age 3 or so.

[whatusername]

Google goggles perhaps? Take a photo of the cool car and have google goggles identify it. I don't think language processing is the answer here -- image/video/audio search (which use those formats as the search term) is probably the way forward.