Comments
Introducing Radiant, a networked forecasting tool for strategic decision-makers
Radiant is a new, standalone app built by Metaculus. It’s a tool for mapping and analyzing the relationships between multiple forecasts at once. It uses this map to help individuals make complex decisions about the future.
By definition, high-stakes decision-makers are tasked with shaping the course of the future. At the outset of their work, they imagine an ambitious goal and consider how likely it is to occur. This prediction is a product of many interrelated predictions about everything from technological breakthroughs to black swan events. Wrestling with these odds and forming a strategy around them requires a tremendous grasp of complexity. However, even the most cunning of scenario planners can only hold so much information in their heads.
If you deal with this type of complex decision-making, Radiant was designed to make your job easier. It weaves the core component of Metaculus – individual forecasts – into a corresponding map of how a given field unfolds. Many discrete forecasts, events, policies, drivers, and trends are linked by their causal relationships to form a bigger picture of the future. With this map in hand, decision-makers can more confidently chart out a course for guiding their initiative from A to Z.
How does it work? Radiant is anchored by two core pieces of functionality:
Visual mapping of forecast networks. Individual predictions are useful, but they rarely capture the full picture of how the future unfolds. Radiant lets you build visual maps where forecasts become nodes and the relationships between them (dependencies, enabling conditions, externalities) become visible.
AI-assisted generation and synthesis. Radiant uses AI to help you formulate forecast questions and discover connections you might have missed. For people who aren't already expert forecasters, AI bots offer a way to access forecasting capabilities approaching the level of the Metaculus community aggregate. Radiant is the easiest way to interface with these bots (however, it’s still fairly simple to run your own forecasting bot, and we encourage you to do so).
Useful interfaces for complex decision-making are critical infrastructure for improving our world. Globally, we spend roughly $3 trillion each year on research and development. These investments rest on assumptions (often implicit and unexamined) about what's technically feasible, when breakthroughs might arrive, and how different fields might converge.
R&D is just one arena. Many major institutional decisions rest on the same kind of implicit, unexamined assumptions. When a biotech firm bets on a therapeutic approach, when a policy team designs climate legislation, when an executive strategizes a supply chain – they're all staking capital on a hidden map of assumptions about what's feasible and when.
Most of these maps of assumptions stay locked in individual heads, never articulated clearly enough to be examined, challenged, or improved. The result is inefficient funding, missed opportunities, and costly surprises in labs, legislatures, and boardrooms.
Radiant is currently in prototype. The core capabilities are live and being refined with early users like Renaissance Philanthropy and ARIA’s Horizon Scanning Study Group (HSSG), who are using Radiant to map out bottlenecks across underexplored research domains like bioengineering and earth-sensing systems. We’re also testing with Metaculus’ consulting clients, who have used Radiant for subjects like AI job displacement and military disempowerment.
Our roadmap includes expanding AI functionality and implementing new features like multi-user collaboration, adversarial collaboration, expected-value calculations, and scenario simulation.
Radiant’s development has been supported in part by generous grants from ARIA and the SFF.
We're inviting a limited cohort of early users to explore Radiant and help shape its development. We're particularly looking to expand our pool of demonstration maps. If you work on policy or complex R&D coordination and want to see what structured forecasting looks like when it's actually usable, we'd love to give you a tour and/or work with you to sketch out a map.
Request a demo: [email protected]
