Question Authority: Make Your Own “Top Tech 2015" Predictions

null
Illustration: iStockphoto

Careful readers will note that the online versions of six articles from IEEE Spectrum’s January “Top Tech 2015” prediction issue offer them a chance to make their own technology forecasts. At the ends of these articles you’ll find boxes with links to questions on robotics, solar power, the Google Lunar XPrize, exascale computers, drones, and smart cars. These lead to SciCast, the prediction market project co-founded by Charles Twardy, Kathryn Laskey, and Robin Hanson at George Mason University.

SciCast lets participants bet points on specific outcomes of questions in science and technology, testing their judgment against thousands of others’ and reaping rewards if their prognostications are better than those of their competitors. Anyone can browse SciCast to see what participants think the future will bring. And anyone can play. Registration is required to make forecasts and join the game, but it’s minimal and free. You just need to pick a username and password. Even an e-mail address is optional. You may be asked for more information, or to complete a simple questionnaire to gauge general scientific and technical literacy, but whether you respond or not is up to you.

The George Mason team built SciCast with support from the U.S. Intelligence Research Projects Activity. It’s one of tools for staying ahead of world events, political and scientific. SciCast focuses on the sci-tech part, and is one of three approaches that have “produced notable results,” according to the Wall Street Journal.

This kind of competition works because it helps eliminate bias and the kind of herd mentality that can seize small groups of experts. As the Journal went on to say:

Deliberation is useful, but it isn't ideal for generating accurate forecasts: It is susceptible to groupthink. Social biases, such as deferring to those with seniority, intrude on the process. And dissenting views often aren't captured. The effects have led analysts to predict events that didn't occur, or miss events that did take place.

In his best-selling The Signal and the Noise: Why So Many Predictions Fail—but Some Don’t, American prognostication guru Nate Silver talked about eliminating bias.

If we want to reduce these biases— we will never be rid of them entirely— we have two fundamental alternatives. One might be thought of as a supply-side approach —creating a market for accurate economic forecasts. The other might be a demand -side alternative: reducing demand for inaccurate and overconfident ones.

A prediction market is the supply-side solution. It lets forecasters work in relative anonymity and isolation, motivated solely by self interest (in theory, at least) as they try to amass the most points. Every bet they make changes the odds: SciCast’s FAQ says that “forecasters spend points to adjust the consensus forecast. Significant changes cost more — but ‘pay’ more if they turn out to be right. So better forecasters gain more points and therefore more influence, improving the accuracy of the system.”

So those out for point-gain and bragging rights can start with these articles, and the associated SciCast questions:

 “When Will We Have an Exascale Supercomputer?,” by Jeremy Hsu

A Robot in the Family,” by Erico Guizzo

The XPrize’s Lunar Deadline Drifts,” by Rachel Courtland 

Big Solar’s Big Surge,” by Peter Fairley 

Flying Selfie Bots,” by David Schnieder

Thus Spoke the Autobahn,” by Philip E. Ross 

Advertisement

Tech Talk

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.

Newsletter Sign Up

Sign up for the Tech Alert newsletter and receive ground-breaking technology and science news from IEEE Spectrum every Thursday.

Advertisement