The most obvious examples given are for ambiguous search queries. When you search for “apple”, do you mean the fruit or the computer? When you search for “cats”, do you want to find the musical, or the animal? Personalization, as it is often outlined, would know you, and be able to complete the missing pieces of the puzzle. It would know you’re a musical lover on holiday, and return Cats the Musical. Sounds worthwhile!
The opposite of this is the search engine (like most today) which doesn’t know you at all. To illustrate this, imagine this conversation between two strangers meeting on a street in London:
Stranger 1: “What time is it?”
Stranger 2: “It’s 7 in New York.”
Stranger 1: “No, I mean here in London.”
Stranger 2: “It’s 4 in some hours.”
Stranger 1: “No, I mean now.”
Stranger 2: “It’s 2.”
Stranger 1: “AM or PM?”
Stranger 2: “Where, here or in New York?"
Nevertheless, this was a trivial example, with trivial problems. While you might have technical means to assign the location to “here” and the local time to “now”, you would still don’t have an answer if Stranger 1 would ask: “What’s the hippest club in town?” (You may ask this your best friend, not so much because he knows you, but because you trust his taste.)
Most personalization problems, thus, are non-trivial. In fact, they often go against the core of what a search engine is trying to solve. This means there are not always technical solutions to these problems. The problems run deeper (and I will avoid the problem of privacy, because who knows – we might be walking into a future where people give up privacy to gain the most from web sites).
This is posing a huge problem to personalization based on past user behavior. Like analyzing the topics most often searched for; the results most often clicked on; or a personal “black list” of sites the user would prefer to avoid. So why isn’t this always good? Because we may not want today what we wanted yesterday.
Just imagine that you would learn to search for information at the age of 8 (today’s generations may start this early), and at this age you love knights and wars and cowboys and games. Would your future search results, like when you’ve turned 16, always circle around these subjects? Imagine how little you would learn. Learning is always getting to know things you do not already know. Analyzing people’s past will only make a search engine understand what they already know, and what they were looking for once. A personalized search engine would risk feeding everyone their own prejudices.
I will only briefly mention this, because it’s a technical problem: for personalization to work, you’d need to provide means for people to login to the site, or find other means to authenticate them. Most people rather take the shortest route to achieving their goal, which would mean logging in is an additional hassle.
I’ve heard someone argue that it would be silly that wen you enter “restaurant”, it doesn’t return only the restaurants close to you. But very often, we’re looking for information on places away from us. Especially when I want to travel to another city the next day will I google for information on this place... like looking for a hotel. Entering “restaurant London” is trivial to the user, and it works globally (for locations near and far). Figuring out where a user wants to eat pizza if same user omits the location is not trivial.
This may be a minor problem, but people like to share search queries. They may tell their friend “Enter ’Starbucks Berlin’ to find out the address of the cafe.” They may also send search URLs pointing to relevant results. People may also want to universally discuss search results, and compare relevancy. A personalized search engine, all of a sudden, would inflict a barrier here.
Let’s say I’d train my search engine to understand what things in life I like. That’s one approach to personalization: I tell my personal Google I like movies, comic books, and ice cream, and I prefer Slashdot to Entertainment Weekly. Which is true.
But you know, I don’t like everything about movies. In fact, I probably dislike most movies. And while it’s true I’m more of a regular at Slashdot than Entertainment Weekly, I’m still bored by Slashdot news on updates to the Linux kernel. Facing the choice, I might even prefer to hear about Britney’s latest scandal.
In fact, because I like something in particular, my preferences in this area are much more refined; this makes me reject more here, not less.
I estimate around 1/4 of my Google searches at work are for others. Say the customer would like you to answer one of his questions. You may not know much about the topic at hand, but you may understand better the power of Google. So you go search for the other person. Personalization would not only be useless in this case, it might be detrimental to the cause. The current search is not about the person performing the search.
This again is not a core problem as it’s a technical issue, to which one can find technical solutions: people switch places all the time. You might access the search engine from work, from home, or (more rarely) from an internet cafe. If the search engine thinks it knows the town I’m in, it’s wrong half of the time: I work in another city than I live in.
Every search engine personalization I would configure in advance would have a big problem: I don’t know in what kind of areas I will be looking for in the future. I might find a new hobby. Or suddenly sell my Apple computer, and get into the fruits business using my Windows PC.
Though it’s fun to play around with Google’s personalization options, I would be incredibly wary to use this in real life. Why? Because I’m no expert on search engine ranking algorithms. Google should stay responsible for providing the best results to me. I can adjust my queries when I know specifically what I want. But I don’t know today what I’m looking for tomorrow, and what kind of settings would be suited best by that time.
Let’s say in my approach to personalize the search engine, I would analyze user data such as age, gender, location or income. In a simple world, this means when an old lady with low income is searching for “race driving”, she’s probably looking for a toy car for her grandchild, and when a young man with high income is looking for “race driving”, he probably wants to buy a real car and actually drive a race.
The truth however is that people aren’t that shallow, and their interests may vary even when this would go against a mass of typical searches for that group. This is related to the problem of the long tail of searching (MSN failed to tackle this and was beaten by Google): you can be good at a small amount of highly common searches (yeah, maybe most older people are not looking to start race driving), but you’re missing out on the large amount of less common searches (the grandfather who wants to spend his retirement money finally living his dreams).
I’d even argue that search queries more often than not delve into the unusual, because we’re always looking for what we don’t know yet. We’re always escaping our immediate context, because we do not need to research our immediate context (we are already familiar with it).
Actually, this is not a problem, it’s the lack of a problem – and yet the biggest issue personalization tries to solve. It’s ambuigity. When I enter “mouse”, do I mean computer mouse, Mickey Mouse, or a pet mouse? Well, why don’t I just tell the search engine? I couldn’t even call my best friend (who knows more than I’d ever want to share with a search engine) and tell him “mouse”, and expect him to know what I want!
Why indeed not just enter “computer mouse” or “Mickey Mouse"? Or “Buy Computer Mouse” or “Mickey Mouse Video"?
This is a general approach which, once learned, always works. And as opposed to many problems associated with personalization, it’s trivial.
>> More posts