M A C H I N E - A S - H U M A N
M E T A P H O R
"If computers can become ill, then they can become healthy. Once healthy, they can think clearly and make decisions."
HE advent of computer technology has been accompanied by the practice of assigning human values to computers. This phenomenon is so subtle that one may not have even noticed it; yet this seemingly harmless practice has ponderous social consequences. When personification is taken too far, the computer can attain attributes which it should not possess, such as accountability. Also, the value of computing technology can be emphasized to the point of undermining the value of human judgment and subjectivity. The future of computing technology, specifically increased dependence on the internet, will greatly promote the machine-as-human metaphor and its accompanying social ills.
In Neil Postman's Technopoly, he describes the machine-as-human metaphor's influence on computers:
Perhaps the most chilling case of how deeply our language is absorbing the "machine as human" metaphor began on November 4, 1988, when the computers around the ARPANET network became sluggish, filled with extraneous data, and then clogged completely. The problem spread fairly quickly to six thousand computers across the United States and overseas. The early hypothesis was that a software program had attached itself to other programs, a situation which is called a "virus." As it happened, the intruder was a self-contained program explicitly designed to disable computers, which is called a "worm." But the technically incorrect term "virus" stuck, no doubt because of its familiarity and its human connections. As Raymond Gozzi, Jr., discovered in his analysis of how the mass media described the event, newspapers noted that the computers were "infected," that the virus was "virulent" and "contagious," that attempts were made to "quarantine" the infected computers, that attempts were also being made to "sterilize" the network, and that programmers hoped to develop a "vaccine" so that computers could be "inoculated" against new attacks.
Postman goes on to explain that the machine-as-human metaphor is not merely a "picturesque anthropomorphism."
"It reflects a profound shift in perception about the relationship of computers to humans. If computers can become ill, then they can become healthy. Once healthy, they can think clearly and make decisions."
This kind of thinking manifests itself in socially-dangerous ways.
The machine-as-human metaphor has two major negative social consequences. The first is an "agentic shift," which encourages society to wrongly associate accountability with computers rather than humans. Postman explains:
"The computer, it is implied, has a will, has intentions, has reasons - which means that humans are relieved of responsibility for the computer's decisions. Through a curious form of grammatical alchemy, the sentence "We use the computer to calculate" comes to mean "The computer calculates." If a computer calculates, then it may decide to miscalculate or not calculate at all. That is what bank tellers mean when they tell you that they cannot say how much money is in your checking account because "the computers are down." The implication, of course, is that no person at the bank is responsible. Computers make mistakes or get tired or become ill. Why blame people? We may call this line of thinking an "agentic shift," a term I borrow from Stanley Milgram to name the process whereby humans transfer responsibility for an outcome from themselves to a more abstract agent. When this happens, we have relinquished control, which in the case of the computer means that we may, without excessive remorse, pursue ill- advised or even inhuman goals because the computer can accomplish them or be imagined to accomplish them."
The second social flaw caused by the machine-as-human metaphor is the undermining of the value of human judgement and subjectivity. The idea that machines can be programmed to possess qualities unique to humans both decreases the value of those qualities and increases our faith in the powers of technical computation. Eventually this kind of thinking leads to the belief that, as Postman explains, "the primary, if not the only, goal of human labor and thought is efficiency; that technical calculation is in all respects superior to human judgement; that in fact human judgement cannot be trusted, because it is plagued by laxity, ambiguity, and unnecessary complexity; that subjectivity is an obstacle to clear thinking."
The Internet's Influence
Since the time the word "virus" was applied to computers, such computer terms have become an accepted part of the English language, solidifying part of the machine-as-human metaphor. The entry for "virus" in the Merriam-Webster Collegiate Dictionary even contains the definition of the term as it applies to computers. Also since this time, the ARPANET has developed into the internet, which promises to further solidify the metaphor to dangerous levels.
As described in the Information Glut essay, the internet has created an ever-expanding sea of information. Before the internet's amount of information reached prohibitive levels, search engines such as Yahoo!, Excite, and Alta Vista allowed users to efficiently find information on the web by extracting web documents which contained a specified set of keywords, such as "venture capitalists Palo Alto."
As the information glut grows, however, the effectiveness of these search engines decreases. Search results often contain thousands of web documents matching the specified keywords but containing no information relevant to the user's interests. Using a different set of keywords is not a solution; the number of results may be reduced but relevant documents will also be excluded. The underlying problem with these search engines is they do not allow the user to precisely define what information he/she is searching for. They require the user to over-simplify a query as a set of keywords, reducing the precision of the query, and causing the search engine to return prohibitive amounts of irrelevant data.
Many internet companies are betting the answer to this problem is the use of "agents," advanced search
tools with user interfaces meant to simulate interaction with an actual person. One example is Jeeves, an
agent representing a butler, who encourages users to define queries using grammatically correct English
A Simple Agent
Jeeves was developed by Ask Jeeves, Inc. , which boasts it is "humanizing the Internet through real-time web interaction."
Jeeves encourages users to define queries using grammatically correct english sentences. After the initial query, he will respond with a list of similar questions for which he knows the answer. The user chooses one of these questions as the final query and the answer is displayed.
For example, asking Jeeves "how do I prevent squirrels from eating my flowers" produces a list of questions including:
Where can I find advice on controlling the garden pest squirrels?
Choosing the first question sends the user to a horticulture document published by the University of Illinois which thoroughly answers the question.
Although Jeeves is a simple agent who has limited interaction with the user, he is still a significant factor in promoting the machine-as-human metaphor. For example, you probably did not notice anything peculiar about Jeeves being described as a "he" and as having the ability to "know" in an earlier paragraph, even though Jeeves is nothing more than a computer program with a clever user interface. This kind of personification is playfully spread throughout the Ask Jeeves web site, but it is taken to extreme levels in more advanced agents which are described as having the ability to "understand, feel, and think."
Extempo Systems, Inc. develops advanced agents, characters which have continuous interaction with users. The agents interact with consumers through "natural-language conversation, animated body language, and electronic actions, such as changing a Web page, running an application, accessing information in a database, or accessing or recording information in a consumer profile," according to a company press release. Extempo's patented Imp Character Technology provides software "minds" for their agents, allowing them to "understand, feel, think, say, and do."
"We humanize the Web experience with interactive characters that provide the same qualities people provide to each other in the real world," says Barbara Hayes-Roth, founder of Extempo Systems. The agents exhibit "job-specific expertise" in roles such as guides, sales assistants, customer service representatives, and librarians. "The characters will also get to know and remember people they meet. They make them feel good," says Hayes-Roth.
After interacting with any of these agents for just a few minutes, it becomes clear that they are trying to
further the machine-as-human metaphor as much as possible. Almost everything the agents say is laced
with emotion or some other human-like quality.
At the Mr. Clean web site, the Mr. Clean agent developed by Extempo Systems helps guide users and collect information about them. Upon starting the Mr. Clean agent, it quickly introduces itself to the user:
"Come talk with me about your cleaning habits, as well as giving me some information about you. This will help me pick out the right coupons, product samples, and other things for your home. I've also started Club Premiere for my most loyal users. If you join this club, you will receive information, coupons, and even samples of new Proctor & Gamble products before they're available in stores."
During the course of a normal conversation, Mr. Clean may say the following things:
"Tell me. what's your first name?"
Although Mr. Clean's conversational skills may still be lacking, it is easy to see how such agents could be a popular tool among Internet users and content providers as well. As the information glut grows and people become more dependent on the Internet for information, these agents will have an increasingly important role and may some day be an essential part of our daily activities.
All of the following technologies already exist in some form, it is just a matter of time before they are applied to agents.
Speech recognition: Instead of typing questions, one will just say, for example, "Jeeves, how do I..."
Voice response: Jeeves will respond in a synthetic but convincing voice, "I recommend..."
Animation: Instead of being a static image, Jeeves will be a 3-dimensional figure whose mouths moves when it talks and whose brow furrows when it "thinks."
Personalization: Users will be able to personalize their agent, specifying things such as what kind of accent it speaks with.
Increased responsibility: Jeeves will not only search the web, but also check your email, and tell you when Ask Jeeves, Inc. stock falls too quickly.
As agents become more advanced, they will become more useful, and people will become more dependent
on them. As agents become more advanced, they will also become more human-like, furthering the
machine-as-human metaphor and intensifying related social problems.
In summary, traditional search tools are becoming unable to usefully extract information from the Internet's
information glut. This problem is giving rise to human-like agents, whose effective operation depends on
solidifying the machine-as-human metaphor. When this metaphor becomes engrained in society's
mind, it spawns destructive ideas about the power and importance of computers. As with the problems of
Internet misinformation and addiciton, the solution to this problem lies solely in the hands of the
individual. One must remember that behind every computer problem lies a responsible human
operator or programmer. More importantly, one must remember technical innovation is important, but equally
important is human judgement, subjectivity, and emotion.