Alignment Problems With Current Forecasting Platforms

TL;DR

This paper identifies and classifies alignment issues in current forecasting platforms, highlighting problems with reward structures and incentives, and proposes potential solutions to improve forecasting accuracy and honesty.

Contribution

It provides a systematic classification of alignment problems in forecasting platforms and suggests specific improvements to scoring rules and incentive mechanisms.

Findings

Current scoring rules are often improper, leading to biased forecasts.

Platforms incentivize strategic behavior that distorts true probabilities.

Proposed solutions aim to align forecaster incentives with truthful reporting.

Abstract

We present alignment problems in current forecasting platforms, such as Good Judgment Open, CSET-Foretell or Metaculus. We classify those problems as either reward specification problems or principal-agent problems, and we propose solutions. For instance, the scoring rule used by Good Judgment Open is not proper, and Metaculus tournaments disincentivize sharing information and incentivize distorting one's true probabilities to maximize the chances of placing in the top few positions which earn a monetary reward. We also point out some partial similarities between the problem of aligning forecasters and the problem of aligning artificial intelligence systems.