The Ethics of the Possible - Chatbotic Sermons

Interesting piece by Deena Prichep, in which clergy agonize as to the ethics of using a chatbot to construct a sermon.

The first point is that it is easy - perhaps too easy. ChatGPT currently advertises its sermon-writing services as follows: 

Your preaching companion. Transform Your Message into Impactful Sermons. Just provide your topic, choose from three tailor-made outlines, and let's co-create a captivating sermon. Fully adaptable to your congregation's needs - denomination, duration, tone, and language.

And for busy clergy the results seem almost touched by the Holy Spirit (aka Ghost in the Machine). Prichep quotes a Lutheran pastor whose first reaction was Oh my God, this is really good. (I may be doing my own research here, but I think there may be something in the Bible about taking the name of the Lord in vain.)

But just because you can doesn't mean you should. One of the arguments in favour of letting a large language model write your sermons for you is that it frees up your time to do more important things, like pastoral care. But are these things really more important? Brad East argues (following Calvin) that the primary task of ministry is the service of Word and sacrament, and that use of Artificial Intelligence shortchanges something essential.

So the underlying principle here seems to be that it might be okay to use AI tools for less important tasks but not for your most important task.

However, there are some other issues with the use of AI tools, including the environmental cost. And East notes the possiblity that large language models might fabricate material as well as pushing a particular agenda, although one might think preachers have always been able to do this without the aid of technology.

---

Brad East, AI Has No Place in the Pulpit (Christianity Today, 27 September 2023)

Deena Prichep, We asked clergy if they use AI to help write sermons. Here's what they said (NPR 17 July 2025) HT Carissa Véliz

Deena Prichep, Encore: Religion and AI, what does it mean when the word of God comes from a chatbot? (NPR 19 July 2025)

John Rector, The Ghost in the Machine (19 June 2024) 

Brad Turner, Beatitudes or Platitudes (Milton Church of Christ, 19 December 2021) 

ChatGPT and the Defecating Duck

For dog owners, the intelligence of dogs shows itself (among other things) in their ability to learn tricks. For cat owners, the intelligence of cats shows itself (among other things) in their disdain for learning tricks. 

When Alan Turing conceived of a way to tell computers and humans apart, now known as the Turing Test, he called it the Imitation Game. His first example was to ask a computer to write poetry - specifically a sonnet on the subject of the Forth Bridge. And his idea of a plausible answer for the computer was to say: Count me out on this one. I never could write poetry.

No doubt many people have tested ChatGPT with exactly the same question. When Jessica Riskin tried it, she was not impressed by its efforts. She found Turing’s imaginary machine’s answer (Turing imitating a machine imitating a human) infinitely more persuasive (as indicator of intelligence) than ChatGPT’s. Turing’s imagined intelligent machine gives off an unmistakable aura of individual personhood, even of charm.

An earlier article by Professor Riskin described a mechanical automaton that attracted large admiring crowds in 18th century Paris. This was a generative pretrained transformer in the shape of a duck, which appeared to convert pellets of food into pellets of excrement. The inventor is careful to say that he wants to show, not just a machine, but a process. But he is equally careful to say that this process is only a partial imitation.

Whereas ChatGPT's bad imitation of poetry is real shit.

---

Jessica Riskin, The Defecating Duck, or, The Ambiguous Origins of Artifical Life (Critical Enquiry, 2003)

Jessica Riskin, A Sort of Buzzing Inside My Head (New York Review of Books, 25 June 2023)

Alan Turing, Computing Machinery and Intelligence (Mind 1950)

The Mad Hatter Works Out

An interesting exchange on Twitter between the mainstream media and the owner of Twitter, which came to my attention via @RMac18 and @karaswisher.

Over a year ago, an American professor wrote a column on MSNBC noting a trend of far right groups using fitness chat groups to recruit and radicalize young men. 

One of the co-founders of Open AI (yes, him) chose to interpret this as asserting that you're a nazi if you work out. There are several possible interpretations of this tweet.

The most unlikely explanation is that a person with a good STEM education and (supposedly) a high IQ has committed a serious error in elementary logic. As in some cats are grey therefore all grey objects are cats.

A slightly more plausible explanation is that the tweet was produced on their behalf by a large language model (LLM), operating a symmetric bi-logic (Matte-Blanco) rather than conforming to classical logic. In the dream world of the unconscious, or in the hallucinations of chat algorithms, the idea that all grey objects are cats might seem perfectly reasonable.

You might just as well say, added the Dormouse, who seemed to be talking in his sleep, that I breathe when I sleep is the same thing as I sleep when I breathe! It is the same thing with you, said the Hatter.

However, the most likely explanation is that the message was deliberately designed to flout logical validity in order to generate the desired affective response - simultaneously appealing to audience A and provoking audience B. (I guess I must be in audience B.) Chasing clicks, as @zsk suggests elsewhere.

Many of the responses adopt similarly dodgy logic, including those that observe (ad hominem) that there are some fat and flabby people on the far right.

---

Arwa Mahdawi, Why is EM borrowing insults from white supremacists? (Guardian, 11 July 2023)

Cynthia Miller-Idriss, Hate in the Homeland: The New Global Far Right (Princeton University Press, 2020)

Cynthia Miller-Idriss, Pandemic fitness trends have gone extreme — literally (MSNBC, 22 March 2022)

For more on LLM and Matte-Blanco, see my post From Chat GPT to Infinite Sets (May 2023)