To be free, information has to be smart (comments on Chris Anderson's "Free!")

WIRED Magazine's editor in chief Chris Anderson, following up on the popularity of his Long Tail meme, theorizes in the March 2008 issue of WIRED about the modern tendency to put information online at no cost. I'll start this blog with the implications of offering free information in the computer field, and build from there to what I agree and disagree with in Anderson's article.

Anderson's taxonomy of “free” contains six models that justify giving the information away. The idea of “free as in freedom” (that is, open source information in the GPL or Creative Commons style) doesn't enter at all into his article. Is that important, given that the article is economic rationale for business? I think it's a crucial omission.

Some lessons from technical computer information

In the case of computer information, Anderson's models for success reduce to the benefit of attracting and keeping users. However you benefit from your computer project–whether by selling the software, charging for service, advertising your skills, or (a motivation more common than usually acknowledged), undercutting somebody else's software–you'll get increasing benefits as the number of users increase. I go into more detail about this in a recent blog.

Another common motivation for improving online information–to reduce service calls and demands on limited developer time–reduces to the goal of streamlining access to information. This is not a motivation for offering more information, but for moving information from expensive resources (expert developers and support staff) to cheap resources (online forums and written documentation).

Online documentation does not currently live up to its potential because computing projects are not smart about online information. Let's reason through the requirements for providing good information.

1. Information has to be smart

Smart documentation, especially for busy users, is all about finding information when they need it. This happens on several levels, going from micro to macro:

* Finding the function or utility needed to solve a specific problem such as “How do I add a drop-cap to a paragraph?” or “How do I extract pages from a PDF?”
* Finding a tutorial on a complicated subject, aimed at the user's particular background and level of expertise.
* Undertaking a field of study that ingrains in users the knowledge they need to use a system wisely, meaning that it meets such long-range goals as security, performance, maintainability, and scalability.

Smart information depends partly on motivating people to write it, partly on ensuring its quality (not just that it's accurate but that it's fit for its users) and partly on helping people find it.

2. Information involves a double problem in motivation

Anderson's article is all about motivation. But technical sites face their own particular motivational challenges. First, they have to motivate developers or other experts to create the information. Then they have to motivate users to read the available resources. Users waste huge amounts of time and may well abandon a project if they find searches frustrating or can't understand the information they find.

The kinds of quality I listed in the previous item requires writers who are emotionally invested in their readers. To get this kind of information, it's not enough to pay developers and technical staff, or to threaten them with guilt or dire consequences for leaving functions undocumented. Viable substitutes for these weak motivations include:

*Ranking authors' participation and quality to build reputation, which they can use to reap rewards inside or outside your project
*Showing authors that their efforts make a difference, by providing feedback from readers (such as on forums), logging hits, providing the kind of reputation system just mentioned, and investing effort to improve their work

These efforts end up increasing readers' motivation as well, by generating more useful documentation and more relevant search results.

3. Smart information requires measurement

To invest time and money in better documentation, projects have to decide where it's worth investing. To start with, logging can identify contributions that are highly sought after. A casual mail message that receives a lot of hits can be extracted and turned into more formal documentation.

At the next level of measurement, ratings show which authors can best meet reader needs. Well-chosen quizzes may be even more informative than ratings, because they prove whether readers got the information they needed. After the project makes an investment–for instance, by hiring someone to rewrite the message–ratings and quizzes can show whether it paid off.

Good search engines provide one type of ranking, but because they distill many kinds of information, they bear only a fuzzy relationship to the needs of a particular reader with a particular background and needs.

4. Smart information develops iteratively and interactively

Here we reach the crux of my argument. Sites that provide instant feedback to authors can improve documentation in several ways.

1. Authors who answer reader questions know in advance that their contributions can potentially make a difference. This draws many more authors than asking them to write a document in the expectation that it might be helpful.
2. Reader responses improve the information by identifying confusing and ambiguous statements. If readers can edit the original posting, they can clear up ambiguities and make the posting more relevant to future readers that come with slightly different needs.
3. A high volume of traffic concerning a particular topic (along with rating systems) identifies which topics are worth investing more effort in, to improve the documentation.
4. The responses, as well as the resulting improvements, reward the authors for making contributions and improve quality.

My argument is that a fluid, open information system meets the goals of the project by generating high-quality information and ultimately happier users. Now, what does this reasoning offer in evaluating Anderson's arguments?

Problems with free

Offering things at no cost has a long history, as Anderson indicates by starting his article with Gillett's canonical “give away the razor” strategy. So far as the Internet goes, Anderson could have rooted his argument more firmly by reminding readers that the Internet infrastructure was free from the start (either downloadable or bundled with the computer).

So current free offerings are more a quantitative than a qualitative shift. I see limits to the trend, though, that Anderson does not discuss. He argues that the costs of providing free information are decreasing, but the true costs of information have never been in the production and distribution of media. They lie in the human intelligence required to produce the information in the first place. Therefore, it's important to look at his sources of funding.

Advertising revenue will decrease

I'm distressed that a new wave of social networking and Web 2.0 sites are being funded through advertising. This is because I'm just making simple deductions (pun intended) from the claim that online advertising is more efficient than ads in broadcast and print media. Whether through Google AdSense, geographic targeting, or other techniques, advertisers are being told they can reach just as many consumers with a real interest in their products through less advertising.

Therefore, unless consumers buy a lot more products, the amount of money spent on advertising throughout the economy will decrease as advertising moves online. Buying more products leads to its own economic and ecological consequences.

Subsidies for information will decrease

Anderson posits that free information can be subsidized by premium versions and sales of related products. For instance, a computer vendor might pay for information about a mobile device in order to sell more copies of the device.

But in a globalized economy, as we all know, anything that can be grown or manufactured can be made in many different places and therefore can be commoditized. (Rising transportation costs will radically change this equation, but will have such a disruptive effect on the ability to sell the goods that I can't consider them here.)

Therefore, decreasing margins on real-world, physical goods will lead to decreasing subsidies for free information.

And if you're hoping to sell premium information–as we do at O'Reilly–your model works only so long as nobody else finds a way to provide something of decent quality for free.

Users are getting fed up with providing information for free

Many of Anderson's models and current Internet sites derive value from contributions by their users, a phenomenon I said back in 2006 calls for more economic research. Increasingly, we read reports that users are wising up. They're asking why companies should be able to make multimillion- and multibillion-dollar deals based on freely provided user information, and give none of the money back to the users. (I'm not even considering the rife privacy issues here.)

The long-term trend of online information may therefore be toward the GPL and Creative Commons form of free information, where users are guaranteed they can benefit from it.

I don't contradict Anderson's assertion that there's a growing phenomenon of free information. Clever ways around the limits I've described may emerge. But I just think that information's current state is highly volatile and that the phenomenon will be driven in very different ways from his six models. “Free as in freedom” may ultimately triumph. Furthermore, professional quality doesn't come for free, so projects and industries have to find ways to fund it.