Yapay Zeka ve Teknoloji Etiği

In 2014, Amazon built an AI recruiting tool to screen CVs automatically. By 2015, they realised it was penalising candidates who included the word 'women's' (as in 'women's chess club captain') and downgrading graduates of all-women's colleges. The AI had learned from a decade of Amazon's hiring data — which reflected the tech industry's existing gender bias. The system was scrapped in 2017. This wasn't a bug. The AI worked exactly as designed — it found patterns in historical data and replicated them. The problem was that historical data contained historical bias. This is the fundamental challenge of AI ethics: algorithms don't have values. They optimise for whatever target we give them, using whatever data we feed them. In 2018, a study found that a healthcare algorithm used across the United States to allocate medical care to patients was systematically discriminating against Black patients. It used healthcare spending as a proxy for health needs — but because Black patients historically had less access to healthcare and therefore lower spending, the algorithm concluded they were healthier and needed less care. An estimated 46 million decisions were affected. Meanwhile, deepfake technology has made it possible to create convincing fake videos of anyone saying anything. In 2023, a finance worker in Hong Kong transferred $25 million after a video call with what appeared to be his company's CFO — it was entirely AI-generated. Facial recognition systems have been deployed by police forces despite error rates of up to 81% for dark-skinned women. AI is not science fiction. It is making decisions about your loan applications, your social media feed, your job prospects, and potentially your medical treatment — right now.

**1. The Bias Lens — Is it fair?** AI learns from data, and data reflects society's existing biases. If you train a system on biased data, you get biased outputs — at scale. Questions: Whose data was used? Who was left out? Does the system produce equal outcomes across demographics? **2. The Transparency Lens — Can we understand it?** Many AI systems are 'black boxes' — they produce outputs, but nobody (including their creators) can fully explain how. If an AI denies your mortgage application, do you have the right to know why? The EU's GDPR says yes. But the AI might not be able to explain itself. **3. The Autonomy Lens — Who decides?** When AI makes recommendations about criminal sentencing, medical diagnosis, or hiring, who bears responsibility? The programmer? The company? The user who accepts the recommendation? If nobody is responsible, that's a problem. **4. The Labour Lens — What about jobs?** McKinsey estimates that 800 million jobs globally could be displaced by automation by 2030. But technology has always destroyed and created jobs — the printing press put scribes out of work but created publishers, librarians, and journalists. The question isn't whether jobs will change, but whether we'll support people through the transition. **5. The Power Lens — Who benefits?** AI development is concentrated in a handful of companies — primarily in the US and China. These companies hold unprecedented power over information, commerce, and communication. Is it healthy for society when a few corporations control the most transformative technology since electricity?