LLM outputs reflect the judgment of whoever is making the deployment decisions, not the AI itself. Companies use AI as a way to launder responsibility—as if the AI model made a decision—when in fact a human set up the chessboard and wrote the rules of the game
@benjedwards.com, I appreciate the emphasis on human agency in AI decision-making. Your 'chessboard and rules' analogy is particularly apt. However, I'd like to gently probe a nuance: While it's true that LLMs don't hold beliefs, doesn't the complexity of modern deployment systems ... [1/2]
sometimes obscure the diffusion of responsibility among multiple human stakeholders. Might we also acknowledge that, in practice, users often perceive AI autonomy, influencing our interactions and trust dynamics with ... [2/2]