Designing for possibility: A sociotechnical deep dive on AI, affordances and the first draft

Technologies offer possibilities, but people, teams and organisational routines shape which possibilities are recognised and used. A sociotechnical perspective helps writers and design teams turn AI from a generic content conveyor into a genuine thinking partner.

Why the same AI behaves differently in different teams

When teams complain that an AI tool produces bland, generic copy, or celebrate the same tool as an insightful thinking partner, they are talking about the sociotechnical system around that AI, not the AI alone. The concept of technological affordances reminds us that tools offer action possibilities; whether those possibilities become real depends on the users, the routines they follow, and the organisational policies and incentives they operate under.

Affordances: action possibilities, not deterministic forces

Gaver’s adaptation of Gibson’s affordance idea, widely used in human–computer interaction, emphasises that affordances are about what an environment allows an actor to do — and that many affordances are perceivable only to some users at certain times. In practice this means a draft suggestion box, a prompt builder or a revision assistant are features that enable certain behaviours; whether a writer sees them as a collaborator, a shortcut, or a threat depends on context and experience.

Adding the sociotechnical lens

The recent work on socio-technical affordances for large-scale collaborations expands this notion: it integrates classic socio-technical systems thinking with affordance theory to explain how tools are mobilised across diverse settings. That research shows the same feature can be harnessed in many ways depending on organisational routines and how teams make sense of the tool’s possibilities. In short: technology doesn’t determine behaviour on its own — the sociotechnical system does.

What shapes whether AI becomes a thinking partner or a conveyor belt?

Perception and training: Do people understand what the tool can do? Per Gaver’s distinctions, affordances can be perceptible, hidden or false. If a useful capability is hidden, it won’t impact practice.
Norms and incentives: If speed is rewarded above reflection, teams will use AI to churn drafts rather than to provoke new ideas.
Routines and roles: Who is expected to draft, critique and revise? Clear roles that include AI-mediated critique can make the system generative.
Feedback loops: Machine learning systems and platform designs create feedback between user habits and the tool’s outputs. That stuff matters for long-term behaviour.
Local infrastructure and context: Research into local sociotechnical affordances shows that global technologies manifest differently in local settings — network, tooling integrations, and regulatory constraints all change what’s possible.

Practical guidance for designers and writing teams

To tilt outcomes towards AI as a thinking partner, treat the tool as one part of a sociotechnical system. The following pragmatic steps help.

Map affordances explicitly. List the tool’s action possibilities (e.g. suggest rewrites, propose outlines, critique logic) and note which are currently perceptible to different roles. Make hidden affordances visible through documentation and demos.
Design rituals that force thinking. Build short routines: a five-minute ‘why this draft exists’ briefing before using AI, or a mandatory human critique step after an AI suggestion. Rituals embed reflective behaviour into the workflow.
Adjust incentives and KPIs. Reward measures that capture thinking quality (e.g. novelty checks, peer review scores) rather than raw throughput. Metrics shape the affordances people act on.
Define roles for the AI. Is it a brainstorming partner, a stylistic editor, or a fact-check assistant? Explicit role definitions reduce misuse and make affordances actionable.
Run small experiments and observe. The socio-technical affordances literature emphasises theorising from practice: run pilot groups, compare how the tool is used, and iterate on routines and training.
Surface local constraints. Consider how local policies, privacy restrictions and tooling integrations change behaviour. Address those constraints when designing the system so affordances remain usable.

Examples of interventions that work

Converting the same AI from a content conveyor to a thinking partner often requires modest organisational design changes. For instance, adding a peer-review gate where AI-generated drafts must be annotated with the human author’s reasoning prompts more explicit thinking. Training sessions that show concrete prompt patterns and the AI’s failure modes turn hidden affordances into perceptible ones. Finally, aligning evaluation metrics to include conceptual novelty or argumentative clarity shifts how people use the tool.

Design checklist for sociotechnical affordances

Have you listed the key action possibilities the AI offers?
Which affordances are hidden or misunderstood by users?
What routines enforce reflection before or after using AI?
Do incentives reward thinking as well as speed?
Are local constraints (privacy, access, integration) accounted for?
Do you have short-cycle experiments to learn how the tool is actually used?

Bottom line

Affordances aren’t magical levers — they are possibilities brought to life inside sociotechnical systems. If you want AI to support thinking rather than replace it, design the surrounding practices, incentives and local infrastructural arrangements so those possibilities are perceived and acted on. The same model can be a collaborator or a conveyor belt; which one it is depends less on code and more on how the system is put together.

Expanded from

This post expands on a point from AI and the First Draft: Liberating Blank‑Page Anxiety or Bypassing the Thinking That Writing Forces?.