This is the subhead for the blog post
OK, this is a bit of a silly post, but for what it’s worth . . .
My wife and I are hooked on watching Who Wants to be a Millionaire (yes, it is still on TV, just not with Regis Philbin). For those of you who have never watched this show, contestants must answer trivia questions with varying degrees of difficulty. If a contestant isn’t sure about an answer, he can use one of three “lifelines” or types of help. The lifelines are:
- Ask the Audience: the user gets the collective opinion of members of the studio audience;
- Phone a Friend: the user calls a friend up and has 30 seconds to get his advice on the question;
- 50/50: The computer removes two out of four incorrect answers, so that the user gets a 50% chance rather than 25% chance.
What does this have to do with search, well think of it this way:
- Ask the Audience: this is the same concept as collaborative filtering or “discovery” search like StumbleUpon, LaunchCast, or Rotten Tomatoes.
- Phone a Friend: Similar to Mahalo, ChaCha – the results are based on a person’s knowledge – humans instead of computers.
- 50/50: Algorithmic search. An algorithm like Google refines results based on rules to come up with the correct answer. Of course, in Millionaire this is a random result, but the point is that a computer is making the selection rather than one or many humans.
The new version of Millionaire, by the way, also has an additional lifeline once the user gets to the $25,000 milestone – the user can “swap out the question” for a new question. You could argue that this is similar to Google’s rarely used “I’m Feeling Lucky” feature.
For what it’s worth, on Millionaire the effectiveness of the lifelines almost always follows this order:
- Ask the Audience is almost always right.
- Phone a Friend is highly variable and is either 100% or next to no value (my sense though is that the most successful phone a friend experiences occur when the friend on the other line just types the question into Google . . .);
- 50/50 is slightly over 50% (since the user has a bit of an idea in advance);
- Swap the Question rarely has any value.
Interestingly – at least in the current state of search – most people would say that the current success rate of different search models would be:
- Algorithmic Search (Google)
- Collaborative Filtering or Discovery
- Human Guides or Directories
- I’m Feeling Lucky
So, in other words, the game show application is almost exactly opposite to the search world application. Of course a game show isn’t the real world, but this does make me think that algorithms aren’t always the best way to find results. Imagine a search engine that truly could ask 100 random people the answer to a search query instantly – it could be pretty powerful. Maybe Millionaire has it right after all?