Part 2 | Putting it to work
Chapter 9. Give it context and constraints
Key points
- Given little information, AI falls back on generalities
- Hand it context and constraints and the output becomes concrete at once
- Therapists are by nature a profession that reads context, which is why the fit is good
Thin input produces generalities
Ask an AI only
"Tell me about approaches to low back pain."
and you will get generalities. Naturally so: from the AI's side, it does not know
- age
- sex
- activity level
- occupation
- living situation
- psychological state
The same is true in the clinic. The word "low back pain" on its own is nowhere near enough information.
Context makes it concrete
Add background:
- woman in her seventies
- chronic low back pain
- high anxiety
- no exercise habit
- living alone
- walks to do the shopping
and the answers become far more specific. This is because generative AI works by
"predicting the next words from context"
The more you tell it about the situation, the better it gets.
Constraints matter just as much
More important still are
constraints
- hard to find an uninterrupted block of time in a day
- lives with family and looks after grandchildren
- travelling next month
- attends once a month
- cannot afford exercise equipment
- has rheumatoid arthritis, so heavy loading is out
However ideal a proposal is, it is worthless if it cannot actually be done. The same is true in the clinic. Tell the AI what the constraints are, and its suggestions become realistic.
The same holds for improving how you work
Not only clinically. When thinking about how to work better:
- about fifteen free minutes at lunch
- five staff
- nobody to run social media
- not good at making materials
- a regional facility, short-handed
- a strong paper culture, little digitisation
Context and constraints change the proposal entirely. "I want to work more efficiently" is abstract; add conditions like the above and you get concrete ideas.
AI is not magic; it is an organising tool
You sometimes hear "I tried AI, but it was underwhelming." One reason is often that
too little information was given to it
AI is not magic; it is a tool that organises what it is given. Adding background, situation, purpose and constraints changes the output a great deal.
Organising the vague
At first you may not even know what is bothering you. That is fine.
- "Work has felt hard lately."
- "There isn't enough time."
- "I want to study but never get anywhere."
Start there. Then, in conversation, work out a little at a time
- what the problem actually is
- what the constraints are
- what you actually want
In a sense, this is putting thought into language.
Reading context is a therapist's real work
A therapist does not look only at symptoms and test results. Support is thought through in terms of daily life, psychology, environment, social background, behaviour and values. That is to say, a therapist is
a professional who reads context
Which is exactly why the habit of "give it context" sits so naturally with a therapist's thinking.
A prompt to try
You are a physical therapist with twenty years of clinical experience.
[Patient]
- woman, 72
- chronic low back pain (over five years); imaging shows mild facet joint arthropathy
- highly anxious, with fear of movement from past experience
- lives alone; shopping is a fifteen-minute walk
- attends every two weeks
- cannot realistically buy exercise equipment
[Constraints]
- the home programme must fit in fifteen minutes a day in total
- it may use only a chair, a towel and a wall, all of which she has at home
- it must not provoke much pain
[Request] Propose three home exercise options for her, each with its purpose, method, repetitions and cautions.
Chapter 9 summary
- Thin input produces generalities
- Hand over context and constraints together and it becomes concrete at once
- The therapist who reads context and the AI think in similar shapes