Introduction

A software agent is a computer program that assists people in information- related tasks, usually across a network like the Internet. Early software agents primarily filtered information and performed repetitive tasks (Maes, 1994). Current software agents not only assist people but can actually stand in for them in the performance of complex information- intensive tasks, such as those agents that buy and sell physical (Maes, 1999) and informational (Flor, Lerch, & Hong, 1999) goods in electronic marketplaces . However, software agents are not just for doing work; they can provide users with entertainment as well.

A social oracle is a particular kind of software agent that produces entertaining social information for a user . The information is typically for amusement purposes only, and not necessarily genuine . Perhaps the earliest example of a social oracle was Eliza (Weizenbaum, 1966) - an agent that simulated a psychoanalyst. Users could ask Eliza questions and it would respond with psychoanalyst-style answers. The enormous growth in online users has driven a concomitant growth in the number of online social oracles. Some of these social oracles, like online horoscopes, are based on popular offline content, e.g., newspaper horoscopes. While other social oracles like Web personality tests, exist offline but are only popular online. Common to social oracles is the ability to take information about a user or a user's friends as input, and output amusing information. Although primarily for user entertainment, businesses can adapt social oracles into effective, low-cost, advertising tools.

Social oracles are part of a growing trend by companies of leveraging game playing for advertising purposes. The most well-known instance of games used for advertising is the 'advergame'-a term coined by Chen and Ringel (2001), which they define as ' interactive gaming technology to deliver embedded advertising messages to consumers by incorporating branding directly into the gaming environment.' It is still too early to know whether advergames actually improve consumers' brand perception or their intent to purchase (Rodgers, 2002). However, the low cost of advergame campaigns makes them an attractive alternative to conventional forms of advertising.

Although social oracles are a form of advertising game, they have a different goal from advergames, per se . Instead of branding, the main goal of a social oracle is to get a user's friends or other associates to visit a business's Web site . Once at the Web site, branding and other forms of conversion can take place, including converting a visitor into a registered customer or getting visitors to purchase a product at the site. Social oracles are intended to be run at the business's Web site by the site's existing community of users. Thus, there are development costs, but no campaign costs. A site's existing user community, through use of the oracles, freely brings in new users. Finally, initial results indicate that the conversion rates for turning visitors into registered users far exceeds the response rates expected from direct marketing, as this chapter reports .

To get a better understanding of why social oracles make good advertising mechanisms, this chapter analyzes two social oracles that were used successfully in building and growing the user community of a special kind of online business known as a programmable autonomous business (Flor, 2003, 2004). This chapter will be structured as follows .

The first section briefly describes: the programmable business that houses the social oracles, the operation of the social oracles, and their conversion rates. The second section is a representational analysis of the social oracles - based on the theoretical framework of distributed cognition (Hutchins, 1995a) - to uncover the social processes that underlie their advertising effectiveness. The analysis will reveal that the social oracles are effective, not because they provide entertaining social information, but rather because they are technological adaptations of ordinary social processes in which people freely exchange information. The third section discusses how technology can automatically transform social information, which is jointly constructed by a user and a social oracle, into an online, direct-response advertisement (Ogilvy, 1985). We end by discussing the implications of this research for more conventional, non-programmable, businesses.

To my knowledge, this chapter presents the first research to examine the re-engineering of basic social processes for advertising purposes. Hopefully, it will serve as a catalyst for future research in this area.

The Programmable Autonomous Business - YesNoMaybe.com (YNM) - and Its Social Oracles

The Web site discussed in this chapter, which contains the two social oracles, is YesNoMaybe.com, or simply YNM . YNM is a Web-based forum developed by Mental Systems, Inc., which has been in operation since February of 2000. It is the most popular teen advice forum on the Web (Alexa, 2004) and largest teen advice forum in the Open Directory ^[1] , (www.dmoz.org) - for both its own category Kids and Teens : Teen Life : Advice : Romance , as well as its parent category - with more than 60,000 registered users who have contributed more than 300,000 postings. YNM's forum consists of 10 different categories, e.g., 'Questions for Girls' and 'Questions for Guys.' Similar to other Web forums, users can: (a) browse a list of topics (Figure 1-1, left screen); (b) read the topic and its opinions (Figure 1-1, right screen); or (c) post their own topics and opinions (see forms at bottom of Figure 1-1). A poll of more than 700 users indicates that 80% are females and 20% are males. Ninety-six percent (96%) of the users are 18 or under, with 62% below the working age of 16.

Figure 1-1: Screen shots: Topic list (left) and topic/opinion display (right)

However, YNM is more than just a teen advice forum, it is also a 'programmable autonomous business': an actual business built entirely in software that runs itself - automatically generating revenue and customers, without any employees or managers guiding it (Flor, 2004). YNM generates revenue by selling advertisement space on its Web pages and has operated profitably since its inception in February of 2000. For example, in its first quarter of operation, when companies had large Internet advertising budgets , YNM averaged $9,500 per month in advertising revenues. Since then revenues have been more modest. Over the past year (9/1/2002-8/31/2003), YNM averaged $153/month in advertising revenues , with only $15/month in expenses (fixed Web hosting fee) - a simply profitability index of 923%. Besides operating profitably for the past three years , YNM continues to grow its user community, averaging 1,031 new user registrations a month over the past year. These users are YNM's customers, without which revenues would cease to exist.

As an autonomous business with neither employees nor managers guiding it, how does YNM continue to grow its customers? One possibility is that users find YNM through search engines. However, aside from its listing in the Open Directory, YNM does not appear within the first ten pages of any of the major search engines, for common search terms like 'teen' or 'teen advice'. Moreover, YNM's owners do not pay to advertise YNM. Thus, YNM must get its customers through other means. In particular, YNM has two social oracles that are adapted for advertising purposes: a Compatibility Game and a 'Love Detective'.

Oracle 1: Compatibility Game, Basic Operation, and Conversion Rate

To play the Compatibility Game, a user clicks on the link labeled 'Love Match?' in the left-hand menu, which brings up the game page. The page instructs the user to enter two names into a form, and to click the 'Love Rating' button. Upon doing so, the game returns a page with three compatibility scores along with a detailed explanation of those scores. The game also gives the user the option of e-mailing the results to one or more friends by entering their e-mail addresses into a text box. Lastly, the e- mails invite the recipients to visit YNM and play the Compatibility Game. Figure 1-2 depicts the Web page for the Compatibility Game (see also Appendix A for a more detailed account of a user playing this game).

Figure 1-2: Oracle 1: Compatibility Game

When a user e-mails a compatibility report to a friend, YNM keeps a record of the e-mail address in a database friend table. This record of e-mail addresses provides a means of assessing the advertising effectiveness of the Compatibility Game, as follows. YNM requires new users to register; the registration form asks for an e-mail address, a secret name , and a password, all of which get stored in a database registration table. If an e- mail address in the friend table matches an e-mail address in the registration table, the Compatibility Game is assumed responsible for that particular registration. During the course of the year spanning December 26, 2001 through December 26, 2002, the Compatibility Game sent out 4,778 unique e-mails. Six hundred and ninety-one of the e- mail addresses in the registration table matched those e-mails sent out by the Compatibility Game - a 14.46% conversion rate ^[2] . This rate is an order of magnitude higher than the 1-1.4% response rate that companies expect from direct mail advertisements (Green & Elgin, 2001). In fact, the oracle's conversion rate may be even higher since most users have multiple e-mail accounts. Thus, a user may receive a compatibility report via one e-mail account, e.g., an AOL e-mail account, yet register on YNM using a different e-mail account, e.g., Hotmail. A simple comparison of e-mail addresses in the friend and registration tables would not count such a user.

Oracle 2: Love Detective, Basic Operation, and Conversion Rate

To play the Love Detective, a user clicks on the link labeled 'Love Detective' in the left- hand menu, which brings up the Web page for the Love Detective. The page instructs the user to enter the e-mail addresses of individuals whom the user suspects have a romantic interest in him or her (Figure 1-3), and then to click the 'Find Out' button. After the user does so, the Love Detective e-mails these suspects and invites them to play the Love Detective as well - to find out who likes them. After a suspect finishes playing, the Love Detective updates a status table (see bottom of Figure 1-3) for both the user and the suspect, which indicates whether or not there was a match: a common e-mail address entered by both the user and suspect (see also Appendix B for a detailed account of a user playing the Love Detective).

Figure 1-3: Oracle 2: Love Detective

Similar to the Compatibility Game, YNM keeps a record of the e-mail addresses used by the Love Detective within a database suspect table, which provides a means of assessing the advertising effectiveness of the Love Detective. To recap, YNM requires new users to register; the registration form asks for an e-mail address, along with other information, all of which get stored in a database registration table. If an e-mail address in the suspect table matches an e-mail address in the registration table, the Love Detective is assumed responsible for that particular registration.

In the time frame spanning December 26, 2001 through December 26, 2002, the Love Detective e-mailed 2,588 unique individuals. Seven hundred and eighty-two of the e-mail addresses in the registration table matched those in the suspect table, indicating a 30.22% conversion rate. One again, this number far exceeds the 1-1.4% response rate expected for direct-mail advertisements. As with the Compatibility Game, the actual conversion rate may be higher, since a suspect may register with a different e-mail address than the one stored in the suspect table.

In short, social oracles are an effective advertising mechanism for YNM, especially compared to direct mail and banner advertisements, which have response rates of 1-1.4% and 0.25% (Nielsen Netratings, 2000), respectively. Note that social oracles' effectiveness figures are based on conversion rates, not response rates, the latter of which would include those users that received an e-mail and visited YNM, but did not register. Gurley (2000) reports that once at a Web site, the conversion rate - the fraction of users that visit a site and either purchase a product or register - is between 3% and 5%. Thus, the social oracles' response rates are almost certainly much higher than their conversion rates.

^[1] The Open Directory is a hierarchical organization of Web sites similar to Yahoo (www.Yahoo.com), but much larger and is outsourced to many of the top search engines, including Google (www.Google.com) and AOL (search.aol.com).

^[2] The fraction of users that visit a site and either purchase a product or fill out a form.