Victor Margolin

Before I go any further, I must confess that I am a man d’un certain age. Therefore my responses to new technology are selective and generational. I still favor transactions with other human beings over those with machines. When I was a boy in the early 1950s, almost all transactions were between one human being and another. The one exception for me was the gumball machine at the candy store around the corner from my apartment building in Washington D.C. I didn’t mind the absence of a live vendor in order to make my purchase. In fact, it was something of a novelty to put a penny in the machine and get a large colored gumball in return.

The gumball machine was only a precusor to the plethora of human-machine transactions that large numbers of people engage in today. In fact, on-line ordering has put many a store out of business and threatened the existence of numerous others. Most people have become accustomed to these transactions and don’t miss the exchange with a live person. However, some vendors and service providers have chosen to simulate a live person on the other end of a telephone in the misguided belief (according to me) that the semblance of such a person will add emotional value to what would otherwise be a bloodless transaction.

Personally, I prefer the basic machine voice to one that simulates the affect of a human. I find it duplicitious to elicit emotional engagement from a customer through a fake human-to-human exchange. Of course, these simulations fool no one and in fact they become the source of considerable frustration when they don’t deliver what they promise. If things are not going well, the simulacrum usually blames the problem on the customer, claiming that he or she did not specify clearly enough what they were trying to accomplish.

Sometimes, if the simulacrum’s frustration reaches a high enough level, it will gracefully bow out of the transaction and turn the customer over to a live operator. However, rather than wait for this moment, I have learned to short circuit the interaction as soon as it begins by saying “Agent,” which is a cue for the machine-human to say, “Ok, I understand you would like to speak with an agent.” Then I get to talk to a real person. As unsatisfying as such machine-human transactions are, the limitations of the desired deception become readily apparent and almost no one is fooled. It would be difficult to mistake the simulated voices for real human ones as they have an awkward cadence that I can’t imagine deceives anyone. Consequently, I don’t expect simulated human voices to ultimately provide any satisfaction, even as artificial intelligence experts expand their capacity for more versatile conversation.

As if it were not enough to simulate human exchanges on the telephone, the use of simulated beings has now moved onto the Internet. I recently received an e-mail from a colleague with whom I was trying to arrange a time to meet for coffee. However, it was not in his name. Instead it bore the moniker of an assistant whose waspy name I have changed to Jane Brown. The e-mail bore her name as the sender as did the return address, although I did not notice that the web address had the term “ai” in it, revealing, though hardly overtly, the program’s identity as one based on artificial intelligence. Since I had never encountered a digital simulacrum presented as a female personal assistant, I did not suspect Jane Brown.

She did, however, appear rather odd and unncessarily formal. The invitation to coffee with my colleague was presented in a visual box similar to the format of various e-vite programs. Inside the box was a specified time for the encounter, 2:00 – 2:45 P.M. I was also informed that my colleague liked to meet at a particular café, which is not the one that I had proposed. I had never encountered such precision when arranging an informal meeting so I decided to query Jane. First, I told her that I would prefer a different café and I specified which one. Then I asked her if the meeting would be limited exactly to forty-five minutes or whether we might run a minute or two over. The response I received came in the form of another box that contained the name of the café I had chosen as well as a new meeting time, which had been extended by forty five minutes so that I was now allotted a period from 2:00 to 3:30 P.M. This was actually longer than I had anticipated spending but I thought better of bothering Jane again since I did not have in mind my own precise duration for the meeting.

I must admit that I found the interaction with Jane quite odd and decided to query my colleague to find out why he encouraged such a strange procedure for arranging a time for coffee. He wrote back and told me that Jane was actually a simulacrum. I can’t say that I was disappointed because she was so proper and emotionally unendearing but I did feel deceived since I had invested a modest measure of my own emotional energy in engaging with her. The deception was furthered by the fact that her name found its place in the lineups of incoming and outgoing e-mails in my mail program, resting comfortably amidst e-mails from real people and overtly mechanical entities. What differentiated my slight encounter with Jane Brown from the intense relationship that Theodore had with Samantha, his OS (operating system) in Spike Jonze’s brilliant film Her, was that I initially believed that Jane Brown was real, while Theodore knew that Samantha was not, even though he was seduced by the depth of her affect.

What Spike Jonze tried to show us with his ending is that Theodore, after Samantha moves on, might be better prepared to have a relationship with a real woman, even though she would likely lack some of Samantha’s unusual qualities. However, with e-mail simulacra presenting themselves as real people, we may engage with them for a time as if they were real, only to feel deceived when and if we find out that they are not. Therefore, what is the point creating a simulacrum as a personal assistant if the objective of the encounter is simply to arrange a meeting or handle a transaction? I am happy enough to enter an on-line exchange with a text that is clearly instructional or aimed at accomplishing something concrete. I have no need for simulated emotional affect and certainly opt for the absence of it. While the technique for simulating a human voice is clear enough, I shudder to imagine what sort of prose would be needed to successfully simulate a live person in an on-line exchange. Would written affect not be as easily detectable as its aural counterpart?

I should also note that the simulated voices in all the telephone transactions I have experienced are those of seemingly white men and women, most likely in their 40s. No older people and certainly no one with an ethnic accent are ever suggested. On the phone, these simulacra have no names but on line their name is part of their identity. Jane Brown is a decidedly wasp name. The nomenclature of simulacra could be potentially explosive, presuming anyone other than a youngish white person would feel discriminated against for not being simulated.

Unlike the machine-like robots that Isaac Asimov wrote about in the 1940s and 1950s or the ones Will Smith encountered in the more recent film I, Robot, the aspirations of some robot scientists is to integrate artificial intelligence and approximate human beings as closely as possible. There is already a market for such humanoids in service industries where the interactions are more complex than those a vending machine can handle. Imagine simulated cosmetologists in Macy’s or waiters and waitresses in Bennigan’s. They will work for nothing and will not require health benefits or pensions. But will we tolerate them as we now tolerate humanoid voices on the phone? I suspect we will. The only saving grace is that we will know they are robots and consequently will save our emotional energy for real people. Should that boundary between the simulated and the real be blurred we will inevitably feel deceived when we are fooled.

Victor Margolin is Professor Emeritus of Design History at the University of Illinois, Chicago. He is currently working on a World History of Design to be published by Berg in London.