We believe that news combined with forecasts of what’s next can build trust and understanding.
It starts with building something better than clickbait news. Yes, clickbaits get page views. But we believe that ethical journalists should avoid this tactic. Here’s why. In a Wharton School of Business research project, Professors Barbara Mellers and Philip Tetlock asked participating college students to forecast on non-controversial topics. The results were conclusive: the participants became more humble and less polarized in their worldviews.
These results are backed up by many research reports, which document similar effects of forecasting on non-controversial topics. For example:
Researchers asked participants to rate the accuracy of a single neutral (nonpolitical, non-COVID-19) headline. Afterwards, participants became better at identifying false news–even when that false news had nothing to do with the neutral topics they were asked to forecast.
Both self-identified Republicans and Democrats improved.
Our Team’s Research Reinforces These Outcomes
Members of the BestWorld team have experienced these outcomes, as well. Our 2019 experiment with a hybrid human-machine forecasting system in IARPA’s Geopolitical Forecasting Challenge II typically resulted in successful predictions around 2/3s of the time. A crucial part of this was that our human forecasters told us the news sources and historical facts they are using to discern what the future holds.
By analyzing approximately 80,000 of their rationales, we found the top two predictors of how well they could foresee geopolitical events were “integrative complexity” and citing similar historical events.
A recent look at forecasting rationales, as well as ongoing research at INFER and our partner Good Judgment, Inc. have revealed yet better ways to forecast by using an AI helper. At INFER, the helper is a specialized version of Anthropic. At Good Judgment, it is a specialized version of ChatGPT.
In 2019-2020, Ai Translate’s chief scientist, Eugene Reyes, together with BestWorld’s president, Carolyn Meinel, figured out how to use his prototype LLM, Semantic Studio, to help her make better forecasts for IARPA’s Geopolitical Forecasting Competition II. The webinar here shows how their forecasting system worked.
BestWorld also builds on forecasting research underway at INFER. At both INFER and Good Judgment, we have found that their forecasters have also built new friendships and communities while reducing polarization. A crucial factor appears to be primarily forecasting on topics that have not been staked out by disinformation or political activism claims. This avoids getting into arguments on topics that can spark highly emotional effects.
Based upon the ten years so far of our team members’ research, we plan to constantly improve and scale up BestWord’s social media/news/forecasting collaborative system. Led by BestWorld chief scientist Dr. Dawna Coutant, we will ensure both transparency and best research ethics as we continue to improve BestWorld.
The research is clear. The effects of combining forecasting with non-controversial topics can fundamentally change how we approach news, social media, and scientific research–driving profound social, political, and interpersonal change.
BestWorld Founding Principles
We seek a more peaceful and prosperous world, one that leaves obody behind. We believe that an integrated forecasting/social media/news aggregation system made safe with the miltilingual moderation system recently created by Linguistic Systems, Inc. could contribute to this goal.
1) Conduct BestWorld so that it encourages unbiased and understanding attitudes toward each other.
That means restisting the temptation to gain members and donations through hate/fear clickbait
2) Conduct forecasting competitions with the goal of doing more than just passively seeing the future. If you and others helping us with forecasts foresee an impending disaster, we will amplify your voice by publicising your insights.
3) We soon will integrate an Activitypub-compliant social media platform with our forecasting system, and join the Fediverse’s over 100 million people through their memberships in Threads, Mastodon, and a fast growing number of others.
4) We also will amplify your insights and forecasts into a news system similar to Flipboard, which now offers users way way to join their
“magazines” to the Fediverse.
.
The Challenge
One of the biggest problems today is the implosion of the traditional news media. If you have ever dreamed of doing a better job of reporting the nrews that what you read online, we can give you better tools than those used in traditional newsrooms. When big news breaks, we’ll arm you with an AI that can help you find people on the scene ready to share their stories with you, and translate if you don’t share the same language.
Our community of seasoned forecasters can assist you to report not just today’s news, but also how your breaking story may unfold over the next days, weeks, and sometimes, even years. Our team already is experimenting with this concept on a small and partial scale at INFER. You are welcome to join us at INFER and have a chance to win prize money.
Our Mission
To offer individuals the best tools to discover and learn with understanding of how to validate knowledge while fostering humility of perception and fidelity of information
BestWorld will bring people, tools, and processes together to help them learn to think critically about the information they’re consuming. Whether the individual grows their own trustworthy knowledge base or learns to identify and rely on accurate informaiton from third-party sources, the goal is the same, depolarization leads the individual to more compassionate views of differing opinions.
Our Vision
To empower individuals to rely on the power of the rational mind to assess information to guide better decision-making
Our plans begin with combining the efforts of a core group of superforecasters with the power advanced large language model (LLM) artificial intelligence. These activities will take place with the guidance of computer scientists, organizational psychologists, and sociologists. Ultimately, the techniques and knowledge we gain during this discovery phase will lead to the development of tools to assess the accuracy of information and the veracity of conclusions relying on that information.
Core Values
Heuristics
Develop a simple, efficient intellectual framework to make near-immediate judgments about the reliability of information, data, or predictions
Humanism
Restore a human-centric focus to complex, data-driven decision-making systems like business intelligence and generative artificial intelligence
Humility
Foster a sense of modesty and acceptance of the limits of our ability to parse complex information, draw conclusions, or make predictions
Sympathetic Responsiveness
Extend to others the grace of understanding their point of view, the power of their biases, and the potential validity of their opinions
The Future of BestWorld
Our next step will be to combine forecasts and our participants’ rationales to produce a news aggregation system. We believe that associating news sources with successful forecasts will build trust in our news aggregation system. This gets around the problem of fact checking services, which ask the reader to simply trust them.
Immediate plans include combining our research with two automated news aggregation and summarization projects at IARPA. The first project we’ll rely on is REASON: Rapid Explanation, Analysis, and Sourcing, a program developing novel technoogoies to assist intelligence analysts tasked with sorting mountains of often-conflicting information. The second IARPA project is BENGAL: BIAS EFFECTS AND NOTABLE GENERATIVE AI LIMITATIONS, aimed at eliminating potential bias from LLM systems. Success on this front woudl be a game changer for the intelligence sector, as well as news and information propagation. Both REASON and BENGAL feed into our own research using leading-edge summarization research and translations of news stories in dozens of languages. Translations will be provided by Linguistic Systems, Inc.
We also will encourage politeness and remove toxic influences with early warnings by both our users and our AI + natural language processing system. A team of humans constantly monitors all systems and our team of journalists. We look to the Good Judgment Open and INFER platforms as models for how to gently and compassionately moderate our systems.
Gentle and Compassionate Moderation
Thank you to our friends there! We also are looking to David Brin’s briefing to Facebook — which Facebook requested, and then ignored! — for ways to gently moderate our participants’ posts.
Ultimately we envision something similar to the IEEE, the world’s largest professional society, which in addition to its massive online presence, hosts thousands of local dinner gatherings, and hundreds of conferences. By building in-person relationships, we will further counter the toxic tendencies of today’s news and social media platforms.
