arXiv subjects

Ville A. Satopää

Publications and source records attributed to Ville A. Satopää.

5 recordsLinked to original sources

arXiv2026

How Proper Scoring Rules Shape LLM Forecasting

This paper evaluates how reward function choice shapes the performance and behavior of LLM forecasters. We compare five proper scoring rules as training objectives for binary forecasts of resolved real-world events. Although the rules share the same theoretical incentive for truthful probability reporting, the resulting models differ in calibration, probability use, and estimated profiles of bias, information, and noise, with smaller differences in aggregate accuracy and discrimination. The Brier-trained model has the lowest observed Brier score and highest AUC-ROC, while the log-trained model has the highest observed log score and lowest calibration error. Models with similar aggregate performance also reach that performance through different combinations of bias, information, and noise. Proper scoring rules therefore need not behave interchangeably as training objectives. Reward choice may shape not only how well an LLM forecasts, but how its forecasting errors are structured.

cs.LG↗

arXiv2018

Combining Information from Multiple Forecasters: Inefficiency of Central Tendency

Even though the forecasting literature agrees that aggregating multiple predictions of some future outcome typically outperforms the individual predictions, there is no general consensus about the right way to do this. Most common aggregators are means, defined loosely as aggregators that always remain between the smallest and largest predictions. Examples include the arithmetic mean, trimmed means, median, mid-range, and many other measures of central tendency. If the forecasters use different information, the aggregator ideally combines their information into a consensus without losing or distorting any of it. An aggregator that achieves this is considered efficient. Unfortunately, our results show that if the forecasters use their information accurately, an aggregator that always remains strictly between the smallest and largest predictions is never efficient in practice. A similar result holds even if the ideal predictions are distorted with random error that is centered at zero. If these noisy predictions are aggregated with a similar notion of centrality, then, under some mild conditions, the aggregator is asymptotically inefficient.

stat.ME↗

arXiv2016

Partial Information Framework: Model-Based Aggregation of Estimates from Diverse Information Sources

Prediction polling is an increasingly popular form of crowdsourcing in which multiple participants estimate the probability or magnitude of some future event. These estimates are then aggregated into a single forecast. Historically, randomness in scientific estimation has been generally assumed to arise from unmeasured factors which are viewed as measurement noise. However, when combining subjective estimates, heterogeneity stemming from differences in the participants' information is often more important than measurement noise. This paper formalizes information diversity as an alternative source of such heterogeneity and introduces a novel modeling framework that is particularly well-suited for prediction polls. A practical specification of this framework is proposed and applied to the task of aggregating probability and point estimates from two real-world prediction polls. In both cases our model outperforms standard measurement-error-based aggregators, hence providing evidence in favor of information diversity being the more important source of heterogeneity.

stat.ME↗

arXiv2015

Modeling Probability Forecasts via Information Diversity

Randomness in scientific estimation is generally assumed to arise from unmeasured or uncontrolled factors. However, when combining subjective probability estimates, heterogeneity stemming from people's cognitive or information diversity is often more important than measurement noise. This paper presents a novel framework that models the heterogeneity arising from experts that use partially overlapping information sources, and applies that model to the task of aggregating the probabilities given by a group of experts who forecast whether an event will occur or not. Our model describes the distribution of information across experts in terms of easily interpretable parameters and shows how the optimal amount of extremizing of the average probability forecast (shifting it closer to its nearest extreme) varies as a function of the experts' information overlap. Our model thus gives a more principled understanding of the historically ad hoc practice of extremizing average forecasts.

stat.ME↗

arXiv2014

Probability aggregation in time-series: Dynamic hierarchical modeling of sparse expert beliefs

Most subjective probability aggregation procedures use a single probability judgment from each expert, even though it is common for experts studying real problems to update their probability estimates over time. This paper advances into unexplored areas of probability aggregation by considering a dynamic context in which experts can update their beliefs at random intervals. The updates occur very infrequently, resulting in a sparse data set that cannot be modeled by standard time-series procedures. In response to the lack of appropriate methodology, this paper presents a hierarchical model that takes into account the expert's level of self-reported expertise and produces aggregate probabilities that are sharp and well calibrated both in- and out-of-sample. The model is demonstrated on a real-world data set that includes over 2300 experts making multiple probability forecasts over two years on different subsets of 166 international political events.

stat.AP↗