iiinova — where your digital self keeps evolving

Reposted by Sebastian Deterding

John Bull @garius.bsky.social · 19/09/2026

tortie peeking out from box with warning on it that says "warning: if left empty and unattended this box will contain a cat

Sebastian Deterding @codingconduct.cc · 02/09/2026

To some extent, we learn in real time by observing what our students with high/low LLM usage fail at. But it’s so hard to establish causality (eg do you fail at x b/c you generally avoid deep thinking which also makes you use LLMs more?)

010

Sebastian Deterding @codingconduct.cc · 02/09/2026

I’d argue that’s what all conscientious educators are trying to figure out atm, and as with employers, are facing massive uncertainty.

110

Sebastian Deterding @codingconduct.cc · 02/09/2026

Aye, I’d say 2/3 of our masters project are LLM-based. In some modules we discourage certain uses since we (& emerging evidence) suggest LLM offloading reduces desirable learning even/especially for an LLM-heavy role. Discerning „good“ to direct a model often needs practice w/o model to develop.

130

Sebastian Deterding @codingconduct.cc · 17/08/2026

Reflecting on the National AI in Research Survey. On substance, any "are you okay with using AI for assessing [X]?" question is a mirror of "are you okay with using a metric for assessing [X]?" We already use - and/or oppose the use of - scalable decontextualised shortcuts to judgement.

000

Sebastian Deterding @codingconduct.cc · 13/08/2026

Mandating a particular design/serendipity algo would be overstep IMO, but I can see EU „safety-test“ standards for assessing diversity/distance measures of the content history of anonymised representative user sample of very large platforms.

000

Sebastian Deterding @codingconduct.cc · 13/08/2026

In fact, yes, Spotify has just announced an AI badge, but I‘s be SHOCKED if tastegraph-based personal generators isn’t deep in development in whatever their AI/innovation unit is.

100

Sebastian Deterding @codingconduct.cc · 13/08/2026

And as platforms vertically integrate to become content houses to increase share of revenue, this becomes even more logical. I.e. when Spotify in-houses AI music generators like Facebook wants to in-house AI people.

100

Sebastian Deterding @codingconduct.cc · 13/08/2026

The content itself may be just as prediction-optimised, but at least the origin is diversified. The principle can be extended to an „indie/art house/…“ ratio like other broadcasting license mandates have it

100

Sebastian Deterding @codingconduct.cc · 13/08/2026

Interesting argument. The one thing I’d add is that maintains exploration/exploitation diversity needs to be a mandate for production and consumption, else we still drift to monoculture. Cf European content mandates forcing streamers to show minimum % regional content.

110

Sebastian Deterding @codingconduct.cc · 05/08/2026

In introspection, I feel the Buddhist viparinama-dukkha is much more on the mark: suffering from the change or passing of something we are attached to by desire. It is immediate pangs of attachment, not interpersonal pressure or coercion.

000

Sebastian Deterding @codingconduct.cc · 05/08/2026

Same for NOT realising an anticipated need satisfaction – i.e., driving by a swing with the buggy and not stopping to swing. Massive emotional frustration. Any anticipated-but-delayed-or-blocked-gratification, really.

100

Sebastian Deterding @codingconduct.cc · 05/08/2026

Again, it's obvious THAT this happens all the time and is strongly aversive. But it's not really need frustration when need satisfaction simply ceases, is it? Would this be framed as autonomy frustration (I WANT TO keep swinging and feel COERCED to stop swinging)?

100

Sebastian Deterding @codingconduct.cc · 05/08/2026

Has anyone in #intrinsicmotivation or #sdt worked on frustration over ceasing intrinsically motivated activity? Like, a toddler screaming bloody murder when they have to leave a swing. It's SO obvious, but I don't recall ANY model or explanation why stopping a fun activity is aversive.

120

Sebastian Deterding @codingconduct.cc · 31/07/2026

But maybe it was useful that a critical mass of people cared regardless, or we unlearn to treat the comms where lack of logic and evidence is CONSEQUENTIAL any different. I vibe-prompt everything and it sounds legit, why not also a government disability allowance decision tree?

030

Sebastian Deterding @codingconduct.cc · 31/07/2026

So much modern org/media communication is transactional void-filling/motion-going with no real need for truth or logic. Maybe LLMs ‚just‘ surface the large share of comms where that was already the case.

130

Sebastian Deterding @codingconduct.cc · 31/07/2026

💯%. It’s what I get as student presentations now, a mock-up argument in coloured boxes. I am getting worried about collective detraining of the ability to spot or care about the absence of real argument.

121

Sebastian Deterding @codingconduct.cc · 31/07/2026

You are not the smartest or the best. You are thrown into a collectively highly smart environment of perspectives and expertise and are able to pull, absorb, and integrate all that input into some forward direction.

000

Sebastian Deterding @codingconduct.cc · 31/07/2026

When they work well, design schools, MBA programmes, incubators and accelerators seem to all embody one idea: that intelligence is collective. That the instigator/creator/founder is a high momentum sponge in a hypertrophic advice environment.

120

Sebastian Deterding @codingconduct.cc · 19/07/2026

Practically, more grateful as a guest, more inquiring and materially prepared as a host when we could (guest towels, blankets, …).

010

Sebastian Deterding @codingconduct.cc · 19/07/2026

Peaked 20s/30s being a host/guest based on living quarters & country. Went down since & focused on a few close, local friends. Further drop to near-0 since child + London living = massive distances. City, space, culture, live stage dictates.

120

Sebastian Deterding @codingconduct.cc · 04/07/2026

Solstice Saturday First steps alone, swings, tantrums Sleepy, hot, and glee #haiku

000

Sebastian Deterding @codingconduct.cc · 02/07/2026

Less than two weeks to apply. If you are a recent/soon-to-be PhD looking to work on games and adolescent wellbeing, do reach out!

010

Sebastian Deterding @codingconduct.cc · 13/06/2026

In a way, the exact mirror image of being a professional creative: doesn't matter the rain, the break-up, the sick dog: you will be able to sit down at 9, and produce art.

010

Sebastian Deterding @codingconduct.cc · 13/06/2026

I feel one part of successful adulting is to be able to compartmentalise space for joy: to be able, in a strict time/space window given by outer demands and under ongoing stress and disaffection, to fully let go into an activity that gives you joy.

120

Sebastian Deterding @codingconduct.cc · 10/06/2026

The moderator’s name is … Vanilla Beer?!?

110

Sebastian Deterding @codingconduct.cc · 08/06/2026

We have 2 job openings – full time, 2 years – on adolescent digital wellbeing and games data in @nballou.bsky.social's Kaleidoscope lab @imperialcollegeldn.bsky.social: * research associate (postdoc): www.imperial.ac.uk/jobs/search-... * research assistant: www.imperial.ac.uk/jobs/search-...

imperial.ac.uk

Description

Please note that job descriptions are not exhaustive, and you may be asked to take on additional duties that align with the key responsibilities ment...

022

Sebastian Deterding @codingconduct.cc · 05/06/2026

* Not all students, some students judiciously off-load mindless/unimportant tasks, YMMV, etc. etc. But the pattern is real.

010

Sebastian Deterding @codingconduct.cc · 05/06/2026

2026 Post-GenAI Student Supervision [Student] Prompts PPT into existence [Supervisor] Critiques PPT [Student] Records and prompts critique to generate new PPT [Repeat] In this loop, the student reduces themselves to a replaceable data entry worker who doesn't run/test their own internal models.

210

Sebastian Deterding @codingconduct.cc · 17/05/2026

Exactly the same scenario. I guide the research and do high level editing (or low level in the first 1-2 papers to show how it’s done), have to take on trust that they did the work but am on the hook if they didn’t.

000

Sebastian Deterding @codingconduct.cc · 17/05/2026

Science is one of MANY systems where the values and best practices we communicate differ widely from actual practice, everyone knows but nobody talks about that gap on the record, and everyone collectively maintaining that tension does essential work to keep the system running as is, POSIWID-style.

000

Sebastian Deterding @codingconduct.cc · 17/05/2026

Once there is an automatic citation resolution checker (does this doi result in a real article page with the same bibliographic details), reviewers & supervisors will use that and we’re back to “cite without reading”

010

Sebastian Deterding @codingconduct.cc · 17/05/2026

Pre-genAI, copy pasting papers from Google Scholar you didn’t read was more cost-benefit efficient, which fit reviewer and supervisor heuristics. This crisis just surfaces they were broken anyway.

120

Sebastian Deterding @codingconduct.cc · 17/05/2026

Eg as supervisor of several PhD students, I CANNOT do my day jobs and verify every citation in every of my students’ papers. I have to rely on setting values, threat, trust, and heuristics (do these citations look legit), even in my domain of expertise.

220

Sebastian Deterding @codingconduct.cc · 17/05/2026

Actually a very common issue. The reality is science has always been built on trust not verification b/c the latter is impossible or cost-prohibitive if done at. Every. Possible. Junction.

120

Sebastian Deterding @codingconduct.cc · 09/05/2026

Jubenshacon: Jubensha is Alibi for socialising/mate-finding with people in a third space.

020

Sebastian Deterding @codingconduct.cc · 09/05/2026

Jubenshacon: here’s one for the Nordic Larpers: Quingganben or emotional scripts. Including player briefings on “how to cry beautifully” during a scene. 360 romance design.

030

Sebastian Deterding @codingconduct.cc · 09/05/2026

And yes, that’s the display screen for an audience of ~150. It’s post-Brexit. What can you do.

020

Sebastian Deterding @codingconduct.cc · 09/05/2026

Jubenshacon: David Middleton comparing Jubensha and escape rooms. Can’t help the TTRPG vibes of “it’s like improv, but on the table with rules.”

120

Sebastian Deterding @codingconduct.cc · 09/05/2026

You have people coming from escape rooms, larp, immersive, murder mystery, TTRPG all going: „how do I play this? Do I roleplay? Solve a puzzle? Go meta or not?“ b/c there is no shared tacit know-how yet. Present gen storygames do a good job here scaffolding first timers with careful scripting

010

Sebastian Deterding @codingconduct.cc · 09/05/2026

Jubenshacon thoughts: this repeats the TTRPG GNS triad as Puzzlism, Dramatism, Narrativism. And the cold start problem of „how to play this?“ when nobody is there yet to model and guide playing culture.

110

Reposted by Sebastian Deterding

Simon Hammann @simonhammann.bsky.social · 20/04/2026

Don't be shy to take on a little two-week side project. These five months will be the most precious three years of your academic journey.

191957537

Sebastian Deterding @codingconduct.cc · 03/04/2026

I thought I was „good at travelling.“ Now with kid and older parents, I realise I was just playing on Easy.

010

Sebastian Deterding @codingconduct.cc · 27/03/2026

To say nothing, of course, of long COVID and all the other damages that will stay with us.

000