Here’s a job robots and AI are unlikely to take over any time soon.
Readers, you may be sick of hearing about the latest occupation to face a threat from AI and/or robots.
When they came for the radiologists etc.
But we bet this application still has some surprise value.
In the traffic of intellectual goods such as ideas, beliefs and norms, the authors of this study on robot preachers propose, the credibility of the purveyor is as important as the content, and nowhere is this more true than in religion.
This seems obvious, yet there are already examples, from as long ago as 2017, of robots being installed in churches of various faiths to deliver sermons, give blessings and even hear confessions.
The authors cite previous work on the “credibility theory of religion”, in which participants need to see their religious elites as both believing and believable in their profession of faith. Mind perception and costly acts of commitment are crucial to this, they write, which puts robots at two disadvantages: “a robot cannot authentically believe in supernatural agents if they do not have the capacity to believe, and they cannot engage in potentially costly behaviour such as celibacy if they are not able to feel the cost”.
In the first of three studies the team went to a temple in Kyoto, Kodai-ji, where the Mindar robot, introduced in 2019, “consistently delivers a 25-min Heart Sutra while turning its head and torso toward different parts of the room”.
Over six weeks they surveyed 422 temple-goers who had seen either Mindar or a human deliver a sermon, measuring perceived credibility and giving participants 1000 yen to keep or donate to the temple as they pleased. Participants who viewed the human found them more credible than those who saw the robot preacher, and were more likely to donate (80% to 68%).
Because this was a natural experiment without randomisation, they continued to a Taoist temple in Singapore where they randomly assigned 239 subjects to watch a human or robot preacher. They measured credibility perception of not only the preacher but the temple, and commitment not only via donations but in willingness to circulate flyers and spread what they had heard.
Exposure to a robot preacher “decreased participants’ perceptions of the preacher’s as well as the temple’s credibility [and] also decreased religious commitment, resulting in fewer donations to the temple and less willingness to spread the message of the sermon”.
But there was still work to do.
Study 3 was an online experiment in which 300 religious participants, Christian this time, were told a sermon was generated either by a human preacher or by an advanced AI. The team asked subjects’ “perceptions of the sermon author’s credibility, mental properties, likeability, charisma, and estimated their religious commitment if they were to hear the sermon at their place of worship”. They also measured anthropomorphism.
The main results were as you’d by now expect. They also found that religious commitment was highly associated with perceived author mind, but this effect was indirect, being fully mediated by author likeability and author credibility.
In an eloquent conclusion, the authors find implications for both automation and for religion, which is in decline around the world: “Domains like religion, which rely on agents modeling their epistemic and moral commitment to belief systems and each other, may not be easily outsourced to robots …
“[O]ne critical factor [in religious adherence] may have been credible religious exemplars who sacrificed their resources, health, and money for their faith. These historical figures draw a sharp contrast to the scandal-prone television evangelists of today and – potentially – the unfeeling robot preachers of tomorrow.”
In the Back Page’s humble opinion, credibility is a sound hypothesis, but we think there may be another aspect to it as well.
As Millicent Weber wrote recently in The Conversation on the subject of creative writing and AI:
We read to enter into a relationship with a story – and through that, with its author. Storytelling and listening are driven by a desire for connection: AI doesn’t complete the circuit.
The Back Page suspects something similar is going on in religion: congregants are seeking not just words of instruction but a kind of relationship (however parasocial or exploitative it may turn out to be) with the priest/rabbi/imam/pandit who is capable of filtering ineffable truths through a human mind just like their own.
A Religibot 3000 won’t complete that circuit either.
Send story tips to penny@medicalrepublic.com.au and we’ll put in a good word.
