AI Ghosts and the Yuck Factor

 

The December Scientific American highlights an article by David Berreby that gets personal.  Berreby's father was born in 1927, the same year as my father, and died in 2013.  Yet the article opens with Berreby asking, "How is your existence these days?" and getting a reply:  "... Being dead is a strange experience."

 

In this conversation, Berreby is using a generative-artificial-intelligence (genAI) version of his father to investigate what it is like to interact with an AI ghost:  a digital simulation of a dead loved one that some psychologists and computer scientists are promoting as an aid for grieving people.

 

I'll be frank about my initial reaction to this idea.  I thought it was terrible.  The "yuck factor" is a phrase popularized by ethicist Leon Kass in describing a gut-level negative reaction to a thing.  He says we should at least pay attention whenever we have such a reaction, because such a sensation may embody wisdom that we can't articulate.  The AI-ghost idea reminded me of Norman Bates, the mentally defective protagonist of Alfred Hitchcock's movie Psycho, who kept his mother around long after her bury-by date and talked with her as though she were still alive. 

 

And to his credit, Berreby admits that there may be dangers to some people whose mental makeup makes them vulnerable to confusing fiction with reality, and could become harmfully addicted to the use of this type of AI.  But in the limited number of cases examined (only 10 in one study) in which grieving patients were encouraged to interact with AI ghosts, they all reported positive outcomes and a better ability to interact with live humans.  As one subject commented, "Society doesn't really like grief."  Who better to discuss your feelings of loss with than an infinitely-patient AI ghost who is both the cause and the solace of one's grief? 

 

Still, it troubles me that AI ghosts could become a widespread way of coping with the death of those we love.  One's worldview context is important here. 

 

Historically, death has been viewed as the portal to the afterlife.  Berreby chose to title his article "Mourning Becomes Electric," a takeoff on Eugene O'Neill's play cycle "Mourning Becomes Electra," which itself was based on the Oresteia play cycle by Aeschylus, a famous Greek playwright who died around 450 B. C.  In the plays, Aeschylus describes the tragic murder of the warrior Agamemnon by his unfaithful wife Clytemnestra, and how gods interacted with humans as things went from bad to worse.  That reference, and a few throwaway lines about ectoplasm and Edison's boast that if there was life after death, he could detect it with a new invention of his, are the only mentions of the possibility that the dead are no longer in existence in any meaningful way.    

 

If you believe that death is the final end of the existence of any given human personality, and you miss interacting with that personality, it only makes sense to use any technical means at your disposal to scratch that itch and conjure up your late father, mother, or Aunt Edna.  Berreby quotes Amy Kurzweil, artist and daughter of famed transhumanist Ray Kurzweil, as saying that we don't usually worry that children will do things like expecting the real Superman to show up in an emergency, because they early learn to distinguish fiction from reality.  And so she isn't concerned that grieving people will start to treat AI ghosts like the real person the machine is intended to simulate.  It's like looking at an old photo or video of a dead person:  there's no confusion, only a stimulus to memory, and nobody complains about keeping photos of our dear departed around.

 

In the context of secular psychology, where the therapeutic goal is to minimize distress and increase the feeling of well-being, anything that moves the needle in that direction is worth doing.  And if studies show that grieving people feel better after extensive chats with custom-designed AI ghosts, then that's all the evidence therapists need that it's a useful thing to do.

 

The article is written in the nearly-universal etsi Deus non daretur style—a Latin phrase meaning roughly "as though God doesn't exist."  And in a secular publication such as Scientific American, this is appropriate, I suppose, though it leaves out the viewpoints of billions of people who believe otherwise.  But what if these billions are right?  That puts a different light on the thing.

 

Even believers in God acknowledge that grieving over the loss of a loved one is an entirely appropriate and natural response.  A couple we have known for 45 years was sitting at the breakfast table last summer praying, and the man suddenly had a massive hemorrhagic stroke, dying later that same day.  It was a terrible shock, and at the funeral there were photos and memorabilia of him to remind those in attendance of what he was like.  But everyone there had a serene confidence that David Jenkins was in the hands of a merciful God.  While it was a sad occasion, there was an undercurrent of bottomless joy that we knew he was enjoying, and that we the mourners participated in by means that cannot be fully expressed in writing.

 

In Christian theology, an idol is something that humans create which takes the place of God.  While frank ancestor worship is practiced by some cultures, and is idolatry by definition, a more subtle temptation to idolatry is offered by AI chatbots of all kinds, and especially by AI ghosts. 

 

While I miss my parents, they died a long time ago (my father was the last to go, in 1984).  I will confess to writing a note to my grandmother once, not long after she died.  So did Richard Feynman write a note to his late wife, who died tragically young of tuberculosis, and a less likely believer in the supernatural would be hard to find. 

 

I suppose it might do no harm for me to cobble up an AI ghost of my father.  But for me, anyway, the temptation to credit it with more existence than it really would have would be strong, and I will take no steps in that direction. 

 

As for people who don't believe in an afterlife, AI ghosts may help them cope for a while with death.  But only the truth will make them free of loss, grieving, and the fear of death once and for all.  And however good an AI ghost gets at imitating the lost reality of a person, it will never be the real thing.

 

Sources:  "Mourning Becomes Electric" by David Berreby appeared on pp. 64-67 of the December 2025 issue of Scientific American.  I referred to Wikipedia articles on "Wisdom of repugnance" and "Oresteia." 

Should Chatbots Replace Government-Worker Phone Banks?

 

The recent slashes in federal-government staffing and funding have drawn the attention of the Distributed AI Research Institute (DAIR), and two of the Institute's members warn of impending disaster if the Department of Governmental Efficiency (DOGE) carries through its stated intention to replace hordes of government workers with AI chatbots.  In the July/August issue of Scientific American, DAIR founder Timnit Gebru, joined by staffer Asmelash Teka Hadgu, decry the current method of applying general-purpose large-language-model AI to the specific task of speech recognition, which would be necessary if one wants to replace the human-staffed phone banks that are at the other end of the telephone numbers for Social Security and the IRS with machines. 

 

The DAIR people give vivid examples of the kinds of things that can go wrong.  They focused on Whisper, which is a speech-recognition feature of OpenAI, and the results of studies by four universities of how well Whisper converted audio files of a person talking into transcribed text.

 

The process of machine transcription has come a long way since the very early days of computers in the 1970s, when I heard Bell Labs' former head of research John R. Pierce say that he doubted speech recognition would ever be computerized.  But anyone who phones a large organization today is likely to deal with some form of automated speech recognition, as well as anyone who has a Siri or other voice-controlled device in the home.  Just last week I was on vacation, and the TV in the room had to be controlled with voice commands.  Simple operations like asking for music or a TV channel are fairly well performed by these systems, but that's not what the DAIR people are worried about.

 

With more complex language, Whisper was shown not only to misunderstand things, but to make up stuff as well that was not in the original audio file at all.  For example, the phrase "two other girls and one lady" in the audio file became after Whisper transcribed it, "two other girls and one lady, um, which were Black." 

 

This is an example of what is charitably called "hallucinating" by AI proponents.  If a human being did something like this, we'd just call it lying, but to lie requires a will and an intellect that chooses a lie rather than the truth.  Not many AI experts want to attribute will and intellect to AI systems, so they default to calling untruths hallucinations.

 

This problem arises, the authors claim, when companies try to develop AI systems that can do everything and train them on huge unedited swaths of the Internet, rather than tailoring the design and training to a specific task, which of course costs more in terms of human input and guidance.  They paint a picture of a dystopian future in which somebody who calls Social Security can't ever talk to a human being, but just gets shunted around among chatbots which misinterpret, misinform, and simply lie about what the speaker said.

 

Both government-staffed interfaces with the public and speech-recognition systems vary greatly in quality.  Most people have encountered at least one or two government workers who are memorable for their surliness and aggressively unhelpful demeanor.  But there are also many such people who go out of their way to pay personal attention to the needs of their clients, and these are the kinds of employees we would miss if they got replaced by chatbots.

 

Elon Musk's brief tenure as head of DOGE is profiled in the June 23 issue of The New Yorker magazine, and the picture that emerges is that of a techie dude roaming around in organizations he and his tech bros didn't understand, causing havoc and basically throwing monkey wrenches into finely-adjusted clock mechanisms.  The only thing that is likely to happen in such cases is that the clock will stop working.  Improvements are not in the picture, not even cost savings in many cases.  As an IRS staffer pointed out, many IRS employees end up bringing in many times their salary's worth of added tax revenue by catching tax evaders.  Firing those people may look like an immediate short-term economy, but in the long term it will cost billions.

 

Now that Musk has left DOGE, the threat of massive-scale replacement of federal customer-service people by chatbots is less than it was.  But we would be remiss in ignoring DAIR's warning that AI systems can be misused or abused by large organizations in a mistaken attempt to save money.

 

In the private sector, there are limits to what harm can be done.  If a business depends on answering phone calls accurately and helpfully, and they install a chatbot that offends every caller, pretty soon that business will not have any more business and will go out of business.  But in the U. S. there's only one Social Security Administration and one Internal Revenue Service, and competition isn't part of that picture. 

 

The Trump administration does seem to want to do some revolutionary things to the way government operates.  But at some level, they are also aware that if they do anything that adversely affects millions of citizens, they will be blamed for it. 

 

So I'm not too concerned that all the local Social Security offices scattered around the country will be shuttered, and one's only alternative will be to call a chatbot which hallucinates by concluding the caller is dead and cuts off his Social Security check.  Along with almost every other politician in the country, Trump recognizes Social Security is a third rail that he touches at his peril. 

 

But that still leaves plenty of room for future abuse of AI by trying to make it do things that people really still do better, and maybe even more economically than computers.  While the immediate threat may have passed from the scene with Musk's departure from DOGE, the tendency is still there.  Let's hope that sensible mid-level managers will prevail against the lightning strikes of DOGE and its ilk, and the needed work of government will go on.

 

Sources:  The article "A Chatbot Dystopian Nightmare" by Asmelash Teka Hadgu and Timnit Gebru appeared in the July/August 2025 Scientific American on pp. 89-90.  I also referred to the article "Move Fast and Break Things" by Benjamin Wallace-Wells on pp. 24-35 of the June 23, 2025 issue of The New Yorker.