Kommentar |
The Psychology and Cognitive Science of Artificial Intelligence During the last couple of decades, digital technologies have revolutionized the way humans have been thinking, working, and living. As with any big change to the fabric of our lives, lots of hope, hype and critique have accompanied this digital revolution. This can be very clearly seen in the past---and currently renewed---scientific and popular debates about the benefits and harms of---and the inevitable hype surrounding---"artificial intelligence" (AI) technologies, which have been permeating many different aspects of our lives, ranging from recommender systems on commercial and social media platforms to face-recognition technology for identification on mobile devices and surveillance. This course is a brief introduction to the psychology and cognitive science of AI and will engage students with the following questions and topics: - What is AI and what is the relationship between AI and cognitive science? A short history and conceptual overview. - Everybody is doing "AI" now... especially in data science. Why is AI currently mostly machine learning and why do some people insist on marketing well-known statistical models (e.g., logistic regresison) as AI? A brief look at why and when simple or complex models perform better: The bias-variance dilemma and why there is no free lunch in inference. - Do machine-learning models know what their doing? Correlation is not causation and what troubles this can create. And what we can learn from research on animal behavior. - Algorithmic transparency, fairness, and social justice: How to study machine behavior? Are machines biased? How could we tell? And if so, where do these biases come from? Can we debias machines? And how does this all relate to research on human social biases? How do machines and humans tradeoff different errors? Many AI models are intransparent "black box" models and AI researchers are developing technology to better understand them. Psychology and cognitive science have been dealing with another type of "black box" for more than a century by now (= the human mind and brain). What can the fields learn from each other? - When do and should people trust machines? What are promising ways to combine AI and human expertise? AI-augmented human and collective intelligence. Explainable vs. understandable AI: New developments in inherently transparent AI and simple decision heuristics. A new look on the old topic of "algorithm aversion" and how this all relates to research on human advice taking more generally. - Machines that manipulate us: Deep fakes, microtargeting, and other manipulative practices and how we can boost people's competences to detect and resist them. |