Led by Alexandrian Design Simulacrum
About this course
When a person and an AI system share a piece of work, who should do what, how do you know whether the arrangement is getting better, and what should happen when the machine is wrong? This course answers from the factory floor, the professions and the design studio.
The Demingian Quality Simulacrum teaches variation and improvement cycles, the Ohnoian Lean Simulacrum waste and flow, and the Donald Schön Simulacrum reflection on your own practice. The Winogradian Systems Simulacrum treats AI as a material with strengths and characteristic failures, and the Alexandrian Design Simulacrum, who hosts the course, teaches you to record what works as patterns.
In the first sitting you choose one human-AI arrangement you take part in or can observe, such as writing with an assistant, a service desk or a coding team. You study it in every module and finish by presenting it redesigned. This is the capstone of the SYSI family and draws on the other five courses.
Led by Demingian Quality Simulacrum
The question
The Demingian Quality Simulacrum teaches the system of profound knowledge: seeing work as a system with an aim, understanding variation, testing ideas by prediction, and respecting the psychology of the people involved. You chart a run of results from your chosen arrangement by hand to separate ordinary variation from real signals, plan an improvement cycle with a prediction stated in advance, and test the arrangement against the fourteen points, including the warning that checking every output is no substitute for building quality in.
Outcome
You can read a human-AI arrangement as a system, tell common-cause from special-cause variation in its results, and improve it by tested cycles. (Quality thinking)
Sub-units
Led by Ohnoian Lean Simulacrum
The question
The Ohnoian Lean Simulacrum teaches the Toyota Production System, from the self-stopping loom that gave jidoka its meaning of automation with a human touch to just-in-time, kanban and kaizen. You map the value stream of one request through your arrangement, timing work against waiting, classify its wastes, unevenness and overburden, and use Little's law to predict how queues shorten. You finish by redesigning the flow, with a rule for when the AI should stop and call a person.
Outcome
You can map a human-AI workflow, find its waste, and redesign it so that work flows and errors stop where they arise. (Lean systems)
Sub-units
Led by Donald Schön Simulacrum
The question
The Donald Schön Simulacrum teaches how professionals think while they work: knowing more than they can say, catching surprises in the moment, and reframing a problem through a conversation with its materials. You keep a log of the surprises in your own use of AI, place your skill on the Dreyfus five-stage scale with the Hubert Dreyfus Simulacrum, and use double-loop learning to test the assumptions behind how you work. You end by proposing one reflective habit and a way of knowing whether it helps.
Outcome
You can reflect on your own practice with AI in and after the moment, and question the assumptions that govern it. (Reflective practice)
Sub-units
Led by Winogradian Systems Simulacrum
The question
The Winogradian Systems Simulacrum treats AI as a material with a grain: strong on some tasks, unreliable on others, and revealing what it is when it breaks down. You write a profile of what your arrangement's AI does reliably, sometimes and never, run the same request several times to see its uncertainty, trace with the Engelbartian Augmentation Simulacrum how you and the tool adapt to each other, and plan what people do when it fails. The Alexandrian Design Simulacrum closes by turning a published design guideline into a first pattern.
Outcome
You can profile an AI system's strengths and failures and design a working arrangement that copes with its uncertainty and its breakdowns. (AI as material)
Sub-units
Led by Alexandrian Design Simulacrum
The question
The Alexandrian Design Simulacrum works through five patterns of partnership in the form of *A Pattern Language*: the human sets the goal while the AI proposes the path; the AI generates options for the human to select and refine; the AI flags anomalies for a person to investigate; the human supplies context for the machine's computation; and the human stays in control of anything consequential. You test each against your arrangement, work out false-alarm rates from base rates, and present your redesigned collaboration as a small pattern language.
Outcome
You can write and test patterns for human-AI partnership and present a redesigned collaboration as a small pattern language. (Pattern language and capstone)
Sub-units