
AI Ethics: 30-Day Growth Guide
Önemli çıkarımlar
- Systems infer far more than you share, so audit what could be guessed from your location, purchases and contacts, not only what you typed.
- When the thing being detected is rare, even an accurate tool flags mostly innocent people, so always ask how common the target is.
- Once a number becomes a target, people and systems start gaming it, so check what an AI tool is actually rewarded for.
- Turn private judgement into written questions for the people who run a system, because institutions move when someone asks on the record.
- Count catches and changes every week, and when a week slips, drop to a five-minute floor instead of restarting the programme.
Bu bit'in içinde
- How the four weeks fit together
This programme runs for four weeks, about twenty minutes a day on five days, with two days kept free for catching up. Each week has one theme, and the themes widen on purpose: week one looks at the data you give and the data others guess, week two at judging a system's claims, week three at how whole systems behave over time, and week four at using your voice with the people who run them.
- Week one: what is inferred, not just shared
Most privacy advice is about what you type. The advanced problem is inference: a system working out things you never said from data you did share.
- Week two: reading accuracy claims like an expert
A claim of ninety-nine per cent accuracy sounds decisive and often is not, because of base rates: how common the thing being detected actually is. When it is rare, a small false-alarm rate applied to everyone else produces far more false alarms than true catches.
- Week three: targets, creep and the blame gap
Week three moves from single decisions to systems over time, where three effects appear. Once a number becomes a target, people improve the number instead of the thing it measured, an effect known as Goodhart's law.
- Week four: from private judgement to public voice
Three weeks of noticing mean little if nothing changes beyond your own settings. Week four turns judgement into voice.
- Measuring progress and rescuing a slipped week
Progress here is measured three ways. Catches should rise, especially in weeks two and three, as you learn to see issues you once walked past.
Altı farklı yolla öğren
Görselleştir, Zihin Haritası, Pratik, Senaryo, Meydan Okuma ve Tekrar, bu kitap için yazıldı.
Test ve bilgi kartları
Okuduklarını test et ve aralıklı tekrar ile aklında tut.
Bits AI
Kitap, bölüm ve o an üzerinde olduğun test sorusu hakkında cevaplar verir.





