Jeremy‘s got an interesting post and discussion on his IR Gupf blog about the process of improving search results. The gist of the discussion is that the process of tweaking search engine performance leads to a local optimum, and cannot be used to discover new interfaces and interaction techniques to support information seeking.
The question raised is how to go about identifying user needs that must be addressed through revolution rather than through evolution of the system design, and the implicit argument is that “you cannot get there from here.”
But this is a problem largely among those whose goals are to optimize ranked retrieval algorithms. But taking ranked retrieval as the essential part of information seeking is a case of premature optimization. Many other strategies, tactics and tools may be used to find information, and the LIS literature describes the broader processes quite well. The problem as I see it is the IR community’s myopia with respect to tools.
This may be symptomatic of Computer Science’s holy grail of trying to build systems that understand what the user wants or needs, and compute solutions to those problems. But at least with respect to exploratory search, such understanding is not only not possible, but also not even desirable in many cases! Exploratory Search is about learning and creating knowledge rather than finding “precompiled” answers in some repository. The reason that the computer cannot answer the searcher’s question (if only it had a good enough algorithm) is that the searcher often doesn’t know what the question is, or that the question evolves over time, and this evolution of understanding is one of the keys to the whole process.
I believe it is more fruitful to conceive of technology that supports information seeking as tools rather than as wizards. We learn how to work with tools through use, through experimentation, through trial and error. But we learn, and we become better craftsmen. We don’t learn anything from wizards, other than that the world is inherently incomprehensible (magic). As systems designers, therefore, we make choices that encourage our users to be either active experimenters or passive consumers. Our society is predicated to a large extent on self-sufficiency and resilience rather than on dependence and helplessness; given the prominence of information in modern life, it seems appropriate to empower people in their acquisition and use of information rather than dumbing-down this valuable process.
FWIW, I caught some flak (offline) from a friend at one of the major web search companies for implying that these companies do think myopically. My friend said that, internally, the discussion and conversations are not as myopic as I think. That may be true, but because the search engines are not transparent in either the tools that they provide, or in the functions that they optimize, it becomes very difficult for me, the user, to know that. And what does it matter, if all I, the consumer, have at the end of the day is the final product — and that product only does the myopic thing (or at least, if it does other things, does so in such a non-transparent manner that I have no way of understanding what is happening well enough to utilize it in the way that I would like)?
I completely agree with you, though, about the distinction between active experimenters and passive consumers. Part of my passion in pursuing information retrieval as a field is that I have a (perhaps mistaken? perhaps too idealistic? perhaps unrealistic?) belief in the notion that IR can improve individuals, and therefore societies. But that can only happen when individuals take personal responsibility, and are (as you say) self-sufficient and actively engaged, rather than dependent and helpless.
I already have television. I thought the Web was supposed to be different.