Getting an AI to summarise the work of 3 AIβs and it only highlights quotes from itself
AI learning office politics then
@acuity.design
Clarity and kindness as foundations for work. Information perception in physical & digital places. Facilitation and workshop design for codesign and civic assemblies. Public prototyping on BlueSky. He/him
Getting an AI to summarise the work of 3 AIβs and it only highlights quotes from itself
AI learning office politics then
If youβre in a βhow to use force on childrenβ meeting then leave
Obama on Jesse Jackson: "The message he sent to a 22 year old child of a single mother with a funny name, an outsider, was that maybe there wasn't any place, any room, where we didn't belong"
Watching BBC Alba documentaries and there must be a queue of film crews filming presenters walking on white sandy beaches of the Outer Hebrides.
βKeep it under your ermine robeβ edition
Enormous hardback book - spine is about 25cm across
1938 Burkes Peerage - the Coronation Honours Special Edition in Oxfam
Did not buy it
Clangers music dvd on shelf in Oxfam
The teens wanted the Clangers OST cd I spotted in Oxfam
On the invention of the mobiiity scooter in 1967 www.bbc.co.uk/sounds/play/... by plumber for his wife who has MS
Very interesting Phd research.cbs.dk/en/publicati... on museums (free to download)
It points how important museums are as spaces for people to be together, to move, to think and to talk. It's less about the artefacts and interpretation as hosting an open space to be curious and social.
Just about to rent Dune part 2 and Amazon (surprisingly) helpfully noted that it comes to Prime in 7 days
'the pupil announced he was βworking on something important about Swanseaβ before lying on the playground tarmac to βfeel the rhythm of the townβ.
...explained to a confused Year 3 class that the sea was βrestless and unforgivingβ.
#WorldBookDay
dailyswansea.wordpress.com/2026/03/05/s...
Games stop
#Gamification doesn't
That's a problem
This is the point made in Wired article last year that AI depend on prompts to anchor their capabilities to a moment of context.
Humans lend emotions, ethics and intents to AI so that they can find stability in swirl of processing
One thing that AI mentioned about IA is that their comprehension is linear.
They cannot glance at a lot of stuff and use tacit knowledge to pick out what meaning matters.
Obviously this is a guide to both clarifying and obfuscating information to AI
As with that WW2 OSI manual on sabotage thru dissentful compliance, there are ways of doing work that doesn't help AI
Make structure shallow Limit hierarchy depth and avoid nested bullets within nested bullets. Keep heading levels consistent and predictable. State relationships explicitly Use direct references (βSee API Rate Limits sectionβ) instead of βsee above.β Restate key assumptions at major transitions. Keep labels and categories consistently Use one sequencing system (Step 1/2/3). Maintain clear category boundaries and stable terminology. Avoid circular or looping definitions Ensure definitions donβt refer back to each other. Keep dependencies directional and easy to follow. Reduce fragmentation Keep related information together. Provide text explanations for tables, diagrams, or visuals.
#IA for #AI checklist
Asked a few AI to say what information architectures they found uncomfortable or incomprehensible.
It's user research for a talk I'm doing at end of March
Asked same IA for AI question to Copilot, ChatGPT and Claude then asked Copilot to summarise.
It has highlighted the points made previously by Copilot in summary
Tony Hancock's The Rebel, a bank of Radiophonic Workshop synths, an orange Mini Cooper driving down the Dr Who time tunnel - and a banging chorus in German. Our Eurovision entry works for me. www.youtube.com/watch?v=niMK...
That the Tories are offering to support Labour on new immigration bill shows how lost the parliamentary parties are. That the parties feel more affinity to each other than any voters.
That an AI can be asked to write a college essay for a student and a target list for a warmonger shows the problem of untethered systems that depends on contextual prompts to anchor them momentarily.
AI is a deeply lost technology. Not on a journey, simply lost at sea and grasping for help.
Writing talk and problem that tacit and implicit mean slightly different things but share the same antonym of explicit
Quite round small brown bird with tuft on top of head standing in morning light in rough grass
Skylark on Rodborough Common yesterday
Tufty bird on tufty grass
Photo by Sue Swain
Gamification is bad because people may not know they are in a game and they donβt know when it is finished.
A game is a game when players have time to stop and step out of the ruleset and appreciate return to standard life.
(Listening to Start The Week)
Move slowly and maintain things - white text on green round badge
Badge rediscovered
Given it was Groupthink that got UK into Iraq, Iβd be happy to discover there was division.
Lack of diversity was what Chilcot Report noted as a strategic problem
A point referenced and βenhancedβ during #BladeRunner
Government-owned train company has been failing on disability awareness training for more than four years
www.disabilitynewsservice.com/government-o...
The structural problem of systems of metric and/or gamification is that there is no clear end point for you to step out and say whether you enjoyed or valued the experience. Listening to Start The Week.
Agentic AI that blocks user agency is, of course, ridiculous product not human centred design/management www.windowscentral.com/microsoft/wi...
Anyway, keep doing really annoying PowerPoints with lots of bullet points, as inconsistently as possible to keep your job.