Jive Talkin'

Is regularly engaging with machines changing the way you talk to people?’

Last week I co-hosted a small monthly networking event at the Royal Society of Arts in The Strand.  The group, a dozen strong, was drawn from members of the Arts and Cultural Network* but the actual composition of the gathering, on the day itself, was decided purely by chance, i.e. whoever had a couple of free hours on that Friday afternoon and was within reach of London’s West End.

The event was scheduled for one o’clock.  By ten past one my co-host and I were getting just a little bit worried. The weather was a balmy 25 degrees, with a pleasant south-westerly breeze along the Thames, and the capital must have been offering a myriad-and-one competing diversions to occupy the creative mind, and…

Don’t panic Mr. Mainwaring!

We needn’t have worried.  One by one our fellow networkers filed in, and each addition to the group added another layer, another unique perspective, to the conversation -so by the time we were ready to close the session we had woven together what was for me (and hopefully for the other participants) a thoroughly engaging and stimulating experience.

Random People Talking

For over a century, The Strand has been the World Heavyweight Champion of Talk. The BBC started broadcasting from Marconi House in 1922, with Bush House transmitting ‘informative, educational and entertaining’ chat around the world, around the clock, seven days a week, until 2012. But as we sat talking on The Strand, addressing a global audience of eleven, I couldn’t help wondering:  

With rocket-heeled AI set on an existential quest ‘To Ubiquity And Back!’  what impact were HAL’s grandkids having on our human-to-human interactions, our conversations and collaborations, dialogue and discourse? 

As we discussed in a previous blog (‘The Fifth Beatle and the T-Shaped Man’ ) AI enables us to acquire a fast and time-efficient grasp of subjects beyond our usual sphere of interest/activity.  And as long as we don’t mistake a smattering of knowledge for total mastery, AI can pave the way for more meaningful conversations with folks ploughing their farrows in different fields from our own; a practice, and process, that almost invariably pays dividends back in our own disciplines.  But aside from engendering this incredibly useful ‘cross-fertilisation,’ there has been precious little work done on how AI may effect our human-to-human interactions.  However what is out there should ring alarm bells, as the chillingly-titled paper ‘Why sycophantic LLMs may imperil interactive norms between humans,’ published last month, suggests:

‘At issue here is not what AI feels (AI systems cannot be offended) but what repeated interaction does to users over time. When a growing portion of communicative life unfolds in contexts where politeness, reciprocity, and conversational repair are neither required nor rewarded, even small efficiency incentives – accumulated across large numbers of daily interactions – could plausibly contribute to gradual shifts in communicative expectations and interactional habits.’

The article cites a number of examples where these ‘shifts’ have been observed, including:

‘The instrumental norm leakage

A growing -though still limited- body of research suggests that interaction with AI can spill over into subsequent human-human judgements and behaviour across multiple domains, including evaluative standards, cooperation, moral enforcement, and interpersonal perception.  Among the clearest causal demonstrations, Tey et al. isolated perceived partner identity as the key driver by having participants interact with the same GPT-4 model, which was presented either as an “AI” or as a purported human (“Alex”).  Those in the AI condition adopted more demanding, instrumental language and displayed reduced positive affect, and subsequently evaluated an unrelated human’s work more harshly, an effect that replicated across both public and private evaluation contexts. Because the downstream evaluation task is distinct from the initial interaction, these findings provide evidence consistent with a cross-context carryover interpretation.’ 

And later:

‘The peril of sycophantic alignment…

...A recent series of studies found that across eleven AI models, chatbots affirmed users’ actions roughly 50% more often than human respondents did -even when users described manipulative or harmful behaviour.  Preregistered experiments further show that even a single interaction with such sycophantic responses can reduce users’ willingness to take responsibility and repair interpersonal conflicts, while increasing their confidence in their own judgements; participants consistently preferred and trusted the more sycophantic systems, potentially creating patterns for such interaction to persist.’ 

The article ends on a stark warning:

‘The goal, simply stated, is to preserve the minimal interactional conditions under which social life remains cooperative and humane, and to help users maintain those conditions in their communicative habits. At stake is not the well-being of AI agents but the psychological residue left on the humans that use them.  If we grow accustomed to commanding articulate, capable entities that rarely resist, never fatigue, and seldom display a standpoint of their own, we may gradually erode our capacity to engage with our fellow humans as equals -the very capacity on which civic life depends.’

I’d like very much like to think that five, fifty, or five hundred years from now, a dozen random strangers will be able to engage as respectfully and as responsively as we did at the RSA last Friday.  However, that may well depend on how acutely we can register, and resist, the more pernicious impacts of AI on our human-to-human interactions.

DT 6 August, 2026

*  The Arts and Cultural Network is a 188,000 strong global community which meets online several times a week.  Visit https://www.linkedin.com/groups/2895827/

** Why sycophantic LLMs may imperil interactive norms between humans Ruolei Gu. Zhyi Chen, Ming Peng, Chao Liu & Zhicheng Lin, Nature Communications Psychology Journal, June, 2026