Workshop Details

Playlist of workshop videos

Following the successful EC 2017 Workshop on Forecasting, we will hold the DIMACS Workshop on Forecasting in 2021. We welcome submissions describing recent research on crowd-sourced, data-driven, or hybrid approaches to forecasting. We especially encourage contributions that leverage forecasts to improve decisions. Please see the Call for Participation below for details.

Recent advances in crowdsourced forecasting mechanisms, including Good Judgment’s superforecasting, prediction markets, wagering mechanisms, and peer-prediction systems, have risen in parallel to advances in machine learning and other data-driven forecasting approaches. Innovations have come from academic researchers, companies, data journalists, and government programs like IARPA’s Aggregative Contingent Estimation program and Hybrid Forecasting Competition.

The workshop will emphasize forecasts embedded inside decision-making systems, where the value of a forecast comes from increasing the expected utility of a key decision. Our ultimate goal is to modernize organizations, markets, and governments by improving how they collect and combine information and make decisions.

The workshop embraces the diversity of this exciting and expanding field and encourages submissions from a rich set of empirical, experimental, and theoretical perspectives. We invite theoretical computer scientists studying algorithmic game theory, incentivized exploration, and NP-hard counting problems; AI researchers studying machine learning, human computation, Bayesian inference, peer prediction, and satisfiability; statisticians studying scoring rules and belief aggregation; economists studying prediction markets, financial markets, and wagering mechanisms; data journalists and marketing scientists studying surveys and polls; blockchain pioneers implementing decentralized prediction markets and other experimental market constructs; social and behavioral scientists studying human behavior modeling; human-computer interaction researchers designing interfaces to facilitate elicitation or convey uncertainty; and practitioners working to improve forecasts as a business or service.

Uncertainty is hard to communicate. Forecasters argue that they are “right”, and critics that forecasters are “wrong” (for example about Brexit or the US Presidential election), despite the fact that probabilistic forecasts can only be evaluated in bulk relative to other forecasts. We invite contributions discussing ways to communicate uncertainty and educate the public about modeling, forecasting, and scoring, building on the excellent 2018 Nova episode “Prediction by the Numbers”.

Topics of interest for the workshop include but are not limited to:

1. Incentives in forecasting. Methods for eliciting truthful and accurate forecasts or information.
2. Coordinating groups of participants to collectively forecast. Examples include prediction markets and wagering mechanisms.
3. Connections between human- and machine-driven forecasting. Uses of data, models, or machine learning in forecasting, and theoretical connections between forecasting mechanisms and machine learning techniques.
4. Making complex forecasts. Predicting structured, combinatorial, or multi-part events. Making conditional forecasts. Forecasting continuous distributions, exponential-sized joint distributions, and spatiotemporal distributions.
5. Forecasting metrics related to climate, the environment, transportation, renewable energy, or public health. For example, metrics of a pandemic including number infected, number hospitalized, number killed, and fatality rate by region and over time, conditioned on public health policies.
6. Forecasting in support of decision making by companies, organizations, or governments.
7. Visualization and other best practices for communicating uncertainty and educating the public about forecasts.

[Videos of workshop talks]  [Short videos for posters]