Artificial Dialectics

Pattern number within this pattern set: 
Warren Sack
UC Berkeley, SIMS

Many online, public discussions can be fairly characterized as "flame wars," or virtual shouting matches; i.e., as violent verbal encounters and not really discussions at all.


Just because the Internet connects people together does not mean that people get along. Putting people in close connection with one another can be a recipe for intimacy, but can also set the stage for violent disagreement. Witness the fate of those subjected to domestic violence.


In 1949 Clad Shannon and Warren Weaver published their mathematical theory of communication. This theory, and subsequent work, has made it possible to understand information and communication technologies (ICTs) as technologies that transmit bits over fixed capacity channels. But isn't there a difference between transmission and communication? Transmission is successful between two people if the receiver can recreate -- perhaps repeat -- the sender's message. In contrast, communication is successful between two people if some form of shared understanding is achieved. Thus, the email message in a my inbox written in a language unknown to me constitutes a successful transmission, but an unsuccessful communication. So -- obviously -- yes, there is a difference between transmission and communication. This difference between transmission and communication explains why -- even as ICTs become more numerous in number and kind -- communication seems to be on the wane. In short, communication technologies -- as they are currently designed -- do not help us communicate with one another.

Even though we can exchange bits with one another via the Internet, we do not necessarily communicate with one another. The ubiquity of online "flame wars" in public discussion forums illustrates the truth of this. In fact, it may very well be the case that the main result of providing universal access to the Internet would be to make it certain that groups fundamentally at odds with one another would butt heads frequently. In the words of the philosopher Jean-Francois Lyotard, online exchanges may very well illustrate a differend, a
difference so vast between participants that it can never be bridged (Lyotard, 1988).

If it is the case, as the Oxford English Dictionary tells us it is, that dialectic was, "originally, the art of reasoning or disputation by question and answer... [and] the regular name of what is now called ‘logic’," then I propose that, in the future, ICTs be designed to facilitate dialectics rather than, as they often do now, simply function to transmit the bits of virtual shouting matches or ad campaigns. The question then becomes the following: How can technologies be designed to facilitate the calm and caring exchange of questions and answers that result in mutual understanding? In other words, can ICTs be transformed into TADs, techne for artificial dialectics?


Recent developments in computational, corpus-based linguistics (e.g., Grefenstette, 1994) make it possible for rough-draft thesauri to be automatically compiled from large collections of texts; e.g., from archives of emails exchanged in online discussions. These algorithms of computational linguistics work in a strictly mechanical manner: they sum and then average together the language of the texts analyzed and, therefore, also average together the words of conflicting authors if the collection analyzed is a collection of messages from a vicious, online exchange. Prototype systems that employ such algorithms to graphically summarize online discussions exist. The Conversation Map (Sack, 2001) is an example of such a system. The descriptive image submitted with this pattern is an example output from the Conversation Map system. It was generated from several hundred email messages exchanged in a heated argument on the Usenet newsgroup soc.culture.palestine. Connected terms in the output diagram are possible synonyms (based on their uses in the texts of the messages). Thus, the machine-generated summarization of the email texts exchanged resulted in a proposal that, for example, "israelis" and "palestinians" are comparable terms; and, also, that "jews" and "arabs" are possible synonyms.

In summary, these algorithms of computational linguistics are doggedly dialectical. Because of the way they are built, they cannot not find a common ground, a synthesis of the language input. Consequently, even for an argument so vicious or incoherent that a skilled, human negotiator might find no place to start building common ground, these algorithms will construct -- through mechanical operations -- a potential synthesis.

Therefore, it is proposed that we begin to design ICTs that facilitate communication by constructing machinery that can identify possible summaries or syntheses of conflicting -- even violently opposed -- messages. While the machine-generated syntheses of today may be unrealistic or naive one can imagine a future generation of ICTs that propose compelling and plausible syntheses that could be used by participants as goals for conversation, as common ground that might be achieved through hard work and sincere discussion.

Pattern status: