Special Feature ~ Large Language Models and Commentaries
This issue opens with Hossein Hassani and Emmanuel Sirimal Silva’s special feature, which explores large language models (LLMs) as potential benchmarks in forecasting practice. On the same subject, Stephan Kolassa exposes the untrustworthiness of LLM forecasts. Christoph Bergmeir provides a commentary questioning whether LLMs can provide good forecasts.
Comments on Foresight Issue 74’s Special Feature ~ Forecasting Performance Objectives
Malte Tichy provided the previous issue’s lead article “Forecasting Performance Objectives: Scaling-Aware Forecast Rating,” noting the unavoidable velocity-dependence of forecast accuracy evaluation, which was followed by three commentaries. In this issue, Malte replies to the commentaries with a plea to use the Ranked Probability Score to evaluate probabilistic forecasts.
Forecast Evaluation
In his article “Unethical Behavior in Forecast Accuracy Evaluation,” Kevin Forbes points out a fundamental flaw in the way renewable energy forecasters calculate forecast error metrics: the practice of creating an “illusion of predictability.” Ken argues misleading forecast evaluation not only tarnishes forecasters’ reputations; it can imperil the world’s critical transition to renewable energy.
Forecasting Competitions
In Matthew Schneider, Jethro Browell, and Rufus Rankin’s piece “Limiting Extreme Behavior in Forecasting Competitions,” restructuring incentives to limit extreme behavior during forecasting competitions is explored.
Tutorial
Ken Fordyce discusses logistic regression for binary outcomes in Part 4 of his tutorial series “Linear Regression with a Time Series View.”
Review
Ira Sohn reviews a report by the research institution Resources for the Future (by Daniel Raimi, Yuqi Zhu, Richard G. Newell, and Brian C. Preston) on the 2024 global energy outlook.
Minitutorial
Stephan Kolassa gives a minitutorial on the Brier Score and its decomposition.
To receive this issue, start your membership to the IIF today! Join Now
