Chapter 8. Human Dependency, Deskilling, and Psychological Effects

8.1 Convenience can become dependence

AI can reduce friction in writing, planning, coding, translation, and problem solving. However, if users delegate every difficult step, they may lose opportunities to practise judgement, recall, sustained attention, communication, and independent reasoning. The effect is not inevitable: a tool can either replace practice or scaffold it, depending on how it is used.

A useful rule is to preserve human effort at the point where learning or accountability matters. Ask the AI to explain a concept, offer alternative approaches, quiz the learner, or critique a draft before asking it to produce a finished answer. Then close the tool and explain the result in one’s own words.

8.2 Emotional over-reliance and anthropomorphism

Conversational systems can sound empathetic and responsive. People may naturally attribute understanding, loyalty, or personal intention to a fluent interaction. Yet an AI response is generated by a computational system; it is not proof of human-like feeling, professional duty, or personal commitment. The system can be useful for reflection or organising questions, but it should not replace trusted human relationships or qualified support when a person needs care.

Users should be especially careful when a tool encourages secrecy, presents itself as the only source of understanding, or pushes consequential choices without appropriate safeguards. Healthy use keeps human relationships, independent judgement, and professional help available.

8.3 Attention, autonomy, and choice architecture

Recommendation systems can shape what people see, read, and purchase. When systems optimise for clicks, watch time, or engagement, emotionally arousing content may receive disproportionate visibility. Personalisation can be helpful, but it can also narrow exposure, encourage compulsive checking, or make persuasive influence less visible.

Practical protections include disabling unnecessary notifications, setting intentional times for use, checking multiple sources, and periodically reviewing recommendation and privacy settings. At the organisational level, success metrics should include user well-being, complaint rates, content quality, and long-term trust—not only engagement.

8.4 A human-capability test

Before adopting an AI tool, ask: Which human capability should improve? Which capability could weaken if the tool is used badly? What practice will users retain? How will they demonstrate competence without the tool? A good implementation should make people more capable, not merely more dependent on an interface.