The good and bad sides of the powerful Semantic Web.
So what happens when the Semantic Web is here? It’s supposed to largely automate the process of searching the Web by allowing us to attach machine-readable assertions (perhaps by using RDF) to information posted on the Web. Then, instead of us poor flailing humans having to painstakingly chase down countless URLs until we get what we want, smart search engines would be able to find precisely what we want in a single shot.
There is an obvious danger to all of this. The new Web will scale, in both good ways and bad. I am certainly not the first person to point out that the smarter the Web, the easier it will be for software to peruse the Web and dig up personal information about us. There will be software that carefully crafts ads in Spam mail that will target our vulnerabilities and our preferences. Websites will dynamically create webpages that target us individually, as well. When we shop online, when we read news, when we make social connections online, the Web will be disarmingly efficient and effective, and this leaves lots of room for fraud and manipulation.
This is already happening to a significant degree, and most of us are aware of it.
The no-longer-hidden database factor.
There is something more subtle about all of this, however. One of the most difficult things to do with traditional Web technology is to expose the content of databases to Web visitors. That’s because the pages that deliver up content pulled from databases are highly dynamic in nature, and so it is very hard for web designers to make search engines (like Google) find and index the content of these databases. There are simple and somewhat effective things web designers can do, like creating static pages that contain terms that are meant to draw web visitors to their sites. These pages are not “destination” pages; rather, they exist only as a way of advertising the information contained in databases.
In the future, RDF assertions (and other machine-readable content) will be added to websites, and they will server as far more effective draws.
But what about privacy? Will web designers inadvertently facilitate fraud and identity theft by enabling the automatic cross-referencing of detailed information existing in databases that have been built and deployed on the Web in isolation? This capability is at the heart of the Semantic Web effort. Information that right now can only be obtained by individual users manipulating individual web interfaces will be discoverable by smart search engines.
The real problem: it will scale.
This is a big deal. It’s not just that previously hidden information will now be discoverable. Because standardized terms and assertions will be used to describe information in databases, smart search engines will be able to automatically interrelate data from otherwise unrelated database systems. When information from multiple places is integrated, new information is effectively created.
For a moment, let’s forget about databases and look at a simple example of information that might be stored statically in two websites. Here is an example adapted from the previous posting of this blog:
Assertion 1: Joe is tall for an athlete.
Assertion 2: Tall athletes should try out for basketball.
A new inference: Joe should try out for basketball.
The point here is that this new inference can be inferred automatically, without the intervention of a human being.
We noted in the previous posting that the information about Joe and the information about basketball might be on different websites. These websites could easily have been built independently. But a key notion – and that is the semantics of the word “tall” in the context of basketball – is what allows this information to be automatically integrated. Another site might point out that Timmy is tall for a kindergarten student, but this would not trigger the suggestion that Timmy try out for the NBA.
Now, let’s get back to database systems, these things that can contain countless terabytes of personal information. Perhaps there is a database at one site containing information about many thousands of athletes. Perhaps there are hundreds or thousands of such sites. The Semantic Web would allow us to find tall athletes without having to know in advance what databases around the world have this sort of data inside them, data that previously could only have been extracted through tedious, time-consume human/computer interaction. Now, a high school counselor or a sports agent looking for new clients can be far more effective at their jobs.
Or, maybe it’s a drug company matching potential customers up with expensive drugs targeted toward specific diseases, or toward people who might have vague symptoms of various diseases, and who might be easily convinced they are sick. Ora con artist looking to scam elderly people who are likely to have dementias.
Or – well, get it? The Semantic Web will scale because it will have access to huge databases, and not just a world wide web of static pages. That’s the danger.