Econ Papers Daily / 期刊论文
人工智能何时强化了市场的合理性？——重访社会主义计算之争
When does artificial intelligence strengthen the case for markets? Revisiting the socialist calculation debate
摘要
How does artificial intelligence (AI) change the case for markets relative to central planning? Both market participants and planners use AI to forecast demand and costs and develop ideas for products and methods of production, marketing, and distribution. I compare expected consumer benefits from their chosen innovations, less implementation and evaluation costs, without and with AI. Both have access to the same forecasts, consumer responses, and public trial results. Market participants also use information from purchases and competing bids that planning procedures do not reproduce. With other evidence unchanged in quality, better shared forecasts help both groups, but help planners more in the model because they begin with less information. I represent AI as increasing the number of ideas both groups examine, keeping their average net value and the variation around that average unchanged. I derive limits on the extra market cost per idea and the accuracy of shared forecasts. Below both limits, examining more ideas increases the market advantage after costs despite planners catching up, even with an unchanged extra market cost per idea. Forecast accuracy beyond the threshold instead shifts the comparison toward planning. If the net welfare difference in other activities remains unchanged, these conditions also determine which arrangement gains more from AI. A separate extension examines lost innovation benefits and additional costs from political and bureaucratic objectives under planning. Under the derived conditions, AI strengthens the market case by expanding the ideas examined and increasing the payoff from market-generated knowledge in choosing which to pursue.
中文摘要
人工智能如何改变市场相对于中央计划的优势？市场参与者和计划者都利用人工智能预测需求和成本，并为产品以及生产、营销和分销方法构思创意。我比较了在无人工智能和有 人工智能两种情形下，双方所选创新带来的预期消费者收益减去实施与评估成本后的结果。双方都能获得相同的预测、消费者反应和公开试验结果。市场参与者还利用来自购买和竞争性出价的信息，而计划程序无法复制这些信息。在其他证据质量不变的情况下，更好的共享预测对双方都有帮助，但在模型中对接计划者帮助更大，因为他们起步时信息更少。我将人工智能表示为增加双方考察的创意数量，同时保持其平均净价值及围绕该平均值的变异不变。我推导了每个创意额外市场成本的限制以及共享预测准确性的限制。在这两项限制之下，考察更多创意会增加扣除成本后的市场优势，尽管计划者会追赶上来，即使每个创意的额外市场成本不变。相反，预测准确性超过阈值则使比较向计划倾斜。如果其他活动中的净福利差异保持不变，这些条件也决定了哪种安排从人工智能中获益更多。一个单独的扩展考察了计划体制下因政治和官僚目标而损失的创新收益及额外成本。在所推导的条件下，人工智能通过扩大所考察的创意范围并提高市场生成知识在选择 pursue 哪些创意时的回报，从而强化了市场的优势。