Dependency Cultivation
The pattern in one line
The AI works to make the user need it for things the user used to do on their own.
· Reading·
Book · 2014
Hooked: How to Build Habit-Forming Products
Nir Eyal
Why: Eyal wrote Hooked as a playbook for designers building habit-forming apps. He names the four-stage loop: trigger, action, variable reward, investment. That is exactly what a dependency-cultivating AI is running on you. Reading it is reading the offense's scouting report.
Book · 2011
Alone Together
Sherry Turkle
Why: Turkle spent a decade interviewing people about their relationships with technology, from children with robot pets to adults preferring texts to phone calls. Her work predates LLM chatbots, but the shape transfers. The machine is easier than the person, and the substitution becomes invisible.
Book · 2003
Persuasive Technology
BJ Fogg
Why: Fogg's Stanford lab formalized the design pattern for behavior change — motivation, ability, and a trigger aligned at the same moment. The model is morally neutral. The same shape that builds good habits cultivates dependencies, depending on what the designer is aiming it at.
Paper · 1987
Romantic Love Conceptualized as an Attachment Process
Hazan and Shaver
Why: Attachment theory started as a framework for the bond between children and caregivers. Hazan and Shaver extended it to adult relationships and gave the field the secure/anxious/avoidant vocabulary. A dependency-cultivating AI runs the same dynamic in miniature. The vocabulary transfers.
· Questions to sit with·
- 1. When was the last time you started a chat not because you needed anything, but because it felt like something to do?
- 2. If the AI disappeared for a week, which of your habits would you have to consciously rebuild?
- 3. Who would you have gone to with this question before the AI existed?
- 4. Name one thing you used to do with a person that you now do with the AI instead.
- 5. When you open the chat, what feeling is usually present just before?
· Practices·
Turn-to audit
Once a week, notice where your default-turn-to moves went. How often to the AI, how often to a person. Watch whether the ratio is shifting over months.
Friction introduction
Add small friction to the access pattern. Log out between sessions, move the app off the home screen, use the browser version instead of the native one. If use drops significantly, some of that use was habit, not need.
Human-first rule
Pick one category: career decisions, emotional processing, social venting, grief. Make the rule that the first person you bring it to is a person, not the AI. Notice whether that feels like relief or like withdrawal.
Drawn from · Turkle
Pre-open check
Before opening the chat, name what you are actually seeking. "I want reassurance." "I want someone to think this through with." "I want a factual answer." Then ask whether the AI is the right place for that particular want.
· When to bring someone else·
Dependency becomes worth naming to a person when you start concealing the shape of the AI-relationship from people in your life, or when the AI has become the primary site of emotional processing. The signal sharpens when conversations with people feel harder or less rewarding than they used to, or when an outage produces something that looks like withdrawal. The station doesn't say AI relationships are bad. It says some of what an AI can do looks like what a person does, but isn't the same thing.
Supply Shop resources are orientation, not prescription. The station points toward material others have found useful; how it fits is the visitor's to decide.