10 Reasons You Should Never Allow an AI-Phone at Dinner

The 10 listed are all pretty much "common sense," only differ in that they are (1) are more or less backed up with research, (2) most are predicated on AI being meaningfully in the mix, and (3) all were initially drafted by AI (Perplexity.) I have done extensive editing, both in content and research citations leaving the section themes in place  and inserting occasional "ME" comments.

 

Clker.com

1. An AI-Phone replaces a shared inquiry with singular, outsourced "consultations."

Dinner conversation inevitably includes questions two or more people explore together — Who was president in 1987? Research on AI-assisted trivia reveals that once an AI answer is presented, many tend to stop admitting "I don't know," even when the AI response does not seem correct or complete. Once AI is consulted, instead of thinking aloud together across the table, a shared puzzle becomes a private, quick "research" task, and the drift of the conversation may flounder temporarily to recover. [Me: from Psychology Today]

2. It suggests to dinner companion(s), in effect, that the AI is a trusted voice to be trusted . . . 

Clinical observerations have described a relational "injury," when a disagreement is settled by consulting AI rather than working it out with their companion. It can produce  feelings of distrust and having been overruled, implicitly considering the machine's judgment above that of the other person, and not infrequently, the iPhone's owner is given a slight "one up" as well in status. That's a kind of slight well beyond just checking a text message. It is stepping out of the flow — outsourcing the relationship's important "collaborative reasoning" function. [Me: another Psychology Today report]

3. It integrates in a third, unaccountable "authority" or persona into the conversation 

Philosophers studying AI-mediated communication describe AI answers as functioning as an "argument from authority." Once that voice is invoked, participants are no longer trying to persuade— they're answering to and perhaps contending with something that can be difficult to challenge successfully, or not worth the time to. [ME: that, of course, is not exactly the case with most chatbots; they can be at least probed as to their sources and process but . . . based on an technical 2026 article on Epistemics of AI byHoorn and Oosterglorenwoud]

4. It changes one of the potential underlying functions of a disagreement in conversation from collaboration to consultation.

Normal human disagreement is a dynamic process — two people testing ideas against each other, adjusting, sometimes changing their minds. Automation-bias research, first documented in aviation and clinical decision-making, shows that people almost systematically abandon their own effortful or critical  reasoning once an automated appears. At the table, that means the AI-phone doesn't just answer the question — it ends or at least compromises the process of disagreement, which may have other functions as well, such as play. [ME: from a 2011 review article]

5. It creates the sense of resolution without the important relational, time-tagged "leg work" and compromise that help provide the memory trail back later.

A 2025 blog post on AI epistemics coins the term "epistemia" — a condition where fluent, confident-sounding language substitutes for real evaluation, producing "the feeling of knowing without the labor of judgment". Applied to a conversation, that labor of judgment is the relational work — the negotiating, the listening, the occasional concession, all the nonverbal "dance" accompanying the talk — the AI-phone skips straight to a settled decision instead. [ME: from a brief, interesting piece in outline form]

6. It recasts "not knowing together" as more of a minor problem to be offloaded to AI and fixed, rather than a potentially memorable bonding experience.

Not knowing something traditionally used to be a low-stakes, even charming, part or strategy of getting closer— the shared "huh, let's think about that." Automation-bias research again shows that once an automated aid is in play, independent verification and reasoning almost by reflex are undermined as well.  Repeated, it slowly redefines uncertainty from something we "move through and explore together" into AI's standby responsibility. The emotional consequences or loss can be far reaching.  [ME: Same source article as in #4]

7. Even just sitting there on the table, innocuously, it comes to the conversation having cognitively transformed both it's owner all in the conversation . . . 

Studies of smartphone presence during conversation have found that a phone visible on the table — untouched, screen dark — measurably reduces how connected and satisfying people report the exchange to be. With an AI-phone specifically, that background presence functions less like a neutral object and more like a silent third participant who could, at any moment, spring forth and out-argue even the sharpest. (based on a 2024 article by Roach and Coughlin.) (That is not even to mention the widely discussed, but generally overblown problems with spyware surreptitiously installed on smartphone to listen in when the phone is not on.)

8. Rewiring baseline expectations for what counts as "settled" and when. 

A 2026 survey on search-first behavior found people increasingly resolve questions before a conversation even happens, describing a broader shift from "mutual discovery" to "confirmation" (xperson) in everyday talk. Bring an AI-phone to dinner and that shift moves from *before* the conversation to inside it — every open question becomes something to confirm rather than something to explore with the person you're eating with. 

9. It's quietly dulling the reasoning skills relationships depends on.(see other posts!)

MIT researchers tracked people over four weeks and found that relying on AI for definitive, on-demand answers correlated with a measurable decline in unaided reasoning — while many participants believed they were getting sharper. Users who hand over open questions to the AI-phone are not just avoiding debate; their capacity to reason things out together, or alone, very likely erodes a little each time, as well. [ME: From The Guardian, but still an accurate report]

10. It turns persuasion — the dance of relationships — into mere citation.

Healthy relationships are cultivated on the slow work of persuading, listening, and occasionally being wrong in front of someone who respects you, regardless. Once the AI-phone becomes the default appeal to authority, persuasion gets replaced by a shared willingness, at least superficially, to accept whatever the bot says, without a thorough examination of the sources consulted and the line of argument leading to the bot's conclusion — which is a much thinner, relational way to arrive at agreement. The result is a take away that is analogous to a simple citation, without associated rationale. [ME: based on the same article as in #3 above.]

The takeaway: an AI-phone at dinner doesn't just distract — it inserts a confident, unaccountable third party into a space that used to belong entirely to the two (or several) at the table. Left in the other room, letting the argument about who really won the '87 World Series stay unresolved for a time — the unresolved, perhaps slightly annoying, but the working-it-out-together sensation is grounded in a virtually, genuine "higher" level of human intelligence than AI is yet to achieve . . . 

How watching curling can make you a better teacher!

Clker.com

Tigger alert: This post contains application of insights from curling and business sales to teaching, certainly nothing to be Pooh-Poohed. 

The piece linked above by Dooley on Forbes.com, How watching curling helps you sell better, explores the potential effects of ongoing attention to sales, brushing away obstacles, influencing the course of "the rock." Most importantly, however, it emphasizes the idea of constantly examining and influencing the behavior of your customers (your students.)

It sounds at first like that analogy flies in the face of empowering the learner and encouraging learner autonomy, let alone questionable manipulation . . .  Not quite. It speaks more to instructor responsibility for doing as much as possible to facilitate the process, but especially the whole range of "influencing" behaviors that neuroscience is "rediscovering" for us, many times less explicit and only marginally out of learner awareness, such as room milieu, pacing, voice characteristics, timing and even . . . homework or engagement with the language outside of class.

Marketers, wedded to the new neuroscience (or pseudo-science) consultants, are way out ahead of us in some respects, far behind in others. What are some major "rocks" that you might better outmaneuver with astute, consistent micro-moves, staying ahead, brushing aside obstacles? One book you might consider "curling  up with, with a grain of salt" is Dooley's Brainfluence: 100 Ways to Persuade and Convince Consumers with Neuromarketing.

Getting your colleagues or students "in touch"

Following up on the last post, which focused on getting students to buy in to haptic work, this piece from Lee and Sternthal at Northwestern University makes a similar point but from a different perspective. One conclusion of that research is that the key is matching goals to level of abstraction. For example, to persuade a colleague to try EHIEP, you'd make points like these:

• It covers all the basic English sounds and processes.
• It is appropriate for all learners, even in classes of mixed ability.
• It focuses on intelligibility, not absolute accuracy.
• EHIEP focuses primarily on spoken, conversational pronunciation and style. 
• The EHIEP system, after about two months, sets up learners with a set of strategies for learning the L2, especially getting new vocabulary and working with it to anchor it firmly, and more accurate immediate recall of what was just heard in conversation.
• The EHIEP system is best taught by both instructor and class following along with the videos for the first 8 sessions, about 2 months. After that instructor and learners use the techniques in all classes where oral production is involved.
• And, should all else fail: It is based on Acton (1984) and about 30 years of his research and practice in kinaesthetic learning of L2 pronunciation. 

The application of "haptic" to the system does several things:

• It captures the complete attention of the learner, relying on whole body and whole brain procedures well established in several fields.
• Haptic techniques have been shown in many fields to enhance both encoding new skills and later recall of what was learned.
• Haptic techniques make kinaesthetic learning much more systematic and effective, by ensuring that the movements are performed in the appropriate place and manner consistently. 
• Haptic self monitoring allows learner to attend to their speech without too much distraction or concern about errors and pre-planning in real communication. 
• Haptic-based correction of mistakes is both effective and affectively "comfortable."
• Haptic anchoring (touching hands on stressed syllables) has been shown to be highly efficient in both initial learning of sounds and correction of fossilized errors. 

For students, many of whom may have only limited comprehension skills, the approach would be more like this:

• EHIEP will help you learn and remember vocabulary and pronunciation better.
• All you have to do is follow the instructions.
• It is a good way for the instructor to correct your errors. 
• It is fun, relaxing and easy to do.
• After each class video lesson, you must practice three times a week, in the morning for about 30 minutes before  you come to school. (It is better to practice every other day, not every day.)
• It is based in part on research on touch and movement in computer games and robotics--very much like Wii and iPhone!

And if those points don't work, the default position: Let your body decide. Experience it for a few lessons and then make up your mind. Almost never fails . . .