Interested in working on this research direction? Apply for our coaching
Want more context on this profile? Explore a map of all our profiles →
This profile is tailored towards students studying economics, philosophy and psychology, however we expect there to be valuable open research questions that could be pursued by students in other disciplines.
Why is this a pressing problem?
The reliability with which we can improve the long-term future depends closely upon how accurately we can predict how future events will unfold. The better we can anticipate the challenges we will face in the future, and how attempts to prepare for these challenges and improve the future will play out long-term, the more likely it is that attempts to improve the world will be effective.
More accurate predictions about the future could support better institutional decision-making across a wide range of domains. For example, a better understanding of the ways in which existential and catastrophic risks could play out in future – for example due to misaligned artificial intelligence or great power war – could allow for more effective policies to be put into place now to mitigate these risks.
However, while the accuracy of short-term forecasting and practices for increasing its accuracy have received significant focus from researchers (for example during the Good Judgement project led by Philip Tetlock), there is currently a lack of research on the accuracy of long-term (≥10yr) forecasts and on the best methods for making these forecasts in different contexts. Consequently, it could be very valuable to better understand the accuracy of our current ability to predict the long-term future and to develop techniques for how this could be improved.
Explore existing research
- Armstrong, Stuart, et al. (2014) The Errors, Insights and Lessons of Famous AI Predictions – and What They Mean for the Future
- Baum, Seth D. (2020) Quantifying the Probability of Existential Catastrophe: A Reply to Beard et al., Futures
- Bernard, David Rhys (2020) Estimating Long-Term Treatment Effects without Long-Term Outcome Data
- Christensen, P., K. Gillingham, & W. Nordhaus (2018) Uncertainty in Forecasts of Long-Run Economic Growth, Proceedings of the National Academy of Sciences
- Grace, Katja, et al. (2018) When Will AI Exceed Human Performance? Evidence from AI Experts
- Granger, Clive & Yongil Jeon (2007) Long-Term Forecasting and Evaluation, International Journal of Forecasting
- Gruetzemacher, Ross, et al. (2021) Forecasting AI Progress: A Research Agenda, Technological Forecasting and Social Change
- Lempert, Robert J., Steven W. Popper, & Steven C. Bankes, Shaping the Next One Hundred Years: New Methods for Quantitative, Long-Term Policy Analysis, RAND Corporation
- Tarsney, Christian (2022) The Epistemic Challenge to Longtermism
For further research, see the GPI research agenda,‘Prediction Markets and Forecasting’ from the Wilberforce Society and this bibliography from the Good Judgement Project.
- The Quantified Uncertainty Research Institute researches practices to help governmental and philanthropic bodies coordinate and make good decisions.
- Making Conversations Smarter, Faster is a project led by Professors Philip Tetlock and Barbara Mellers to develop a collective reasoning system to enhance processes like policy debates or collective forecasting exercises.
- Good Judgement offers training and services relating to forecasting.
- Epoch is a research group working on forecasting the development of transformative AI.
- AI Impacts is a research organisation researching the likely impacts of human-level AI, including by working to understand AI timelines.
- Global Priorities Institute carries out foundational research on how to do the most good. One of the topics in their latest research agenda was forecasting the long-term future.
- The Forecasting Research Institute develops forecasting methods to improve decision-making on high-stakes issues.
Find a thesis topic
If you’re interested in working on this research direction, below are some ideas on what would be valuable to explore further. If you want help refining your research ideas, apply for our coaching!
There are a number of different directions that could be pursued. The Global Priorities Institute suggests various questions in their research agenda relevant to economics, such as ‘how accurate have forecasts about long-term developments been in the past?’ and ‘Is it possible to incentivise accuracy in long-term forecasts, given that we may not live to see the realisation of the event we wish to forecast?’
It could also be useful to explore the key methodological questions described in this report that Philip Tetlock and Pavel Atanasov have identified as needing to be resolved before a global catastrophic risk forecasting tournament can be held.
For work related to forecasting and AGI specifically, ‘Forecasting AI Progress: A Research Agenda’ provides a summary of research questions suggested by experts.
It could also be useful to explore the direction suggested in this paper, which states that ‘a major obstacle to the practical use of prediction methods is the lack of understanding of the environments in which they are accurate. For instance, for prediction markets, research could tackle a wide range of conditions, including market liquidity, time ranges, information availability, or participant selection, in order to evaluate the parameter space in which prediction markets can be used reliably.’ (p.48)
The Global Priorities Institute suggests various questions in their research agenda relevant to philosophy:
- Forecasting the long-term effects of our actions often requires us to make difficult comparisons between complex and messy bodies of competing evidence, a situation Greaves (2016) calls “complex cluelessness”. We must also reckon with our own incomplete awareness, that is, the likelihood that the long-run future will be shaped by events we’ve never considered and perhaps can’t fully imagine. What is the appropriate response to this sort of epistemic situation? For instance, does rationality require us to adopt precise subjective probabilities concerning the very-long-run effects of our actions, imprecise probabilities (and if so, how imprecise?), or some other sort of doxastic state entirely?
- In recent centuries, revolutionary ideas like evolutionary theory and computing have vastly reshaped both our understanding of the physical world and the actual workings of the social world. It is natural to suppose that other similarly transformative ideas will be discovered in coming centuries. Does our ignorance of future transformative ideas constitute a major obstacle to predicting and influencing the far future, as Deutsch (2011) suggests? If so, how should we respond? To what extent, if at all, does this undercut the case for longtermism?
There are a number of different directions that could be pursued. Some questions in the Global Priorities Institute research agenda might be explored by students studying psychology, such as ‘Is it possible to incentivise accuracy in long-term forecasts, given that we may not live to see the realisation of the event we wish to forecast?’
The research agenda ‘Psychology for Effectively Improving the Future’ suggests several research questions related to long-term forecasting, including ‘what scoring rules are most effective for eliciting accurate forecasts?’ and ‘how well do schemes like the Bayesian truth serum work for incentivising truthful forecasting in contexts where the forecast is never resolved?’
It could also be useful to explore the key methodological questions described in this report that Philip Tetlock and Pavel Atanasov have identified as needing to be resolved before a global catastrophic risk forecasting tournament can be held.
For work related to forecasting and AGI specifically, ‘Forecasting AI Progress: A Research Agenda’ provides a summary of research questions suggested by experts, some of which could be addressed by psychology researchers.
Further resources
To learn more, you could start by exploring:
- Research from some of the organisations listed above, for example Ought’s work on automating forecasting, AI Impacts’ research on predicting AI timelines, and research from Rethink Priorities.
- Why I think it’s important to work on AI forecasting from the EA Forum.
If you’re interested in working on this research direction, apply for our coaching and we can connect you with researchers already working in this space, who can help you refine your research ideas.
You can also apply to join our community if you’re interested in peer connections with others working in this area.
Apply for our database of potential supervisors if you’re looking for formal supervision and take a look at our advice on finding a great supervisor for further ideas.
Our funding database can help you find potential sources of funding if you’re a PhD student interested in this research direction.
- Nuno Sempere’s forecasting newsletter covers forecasting news, with a focus on experimental forecasting.
- You can sign up for our newsletter to hear about opportunities such as funding, internships and research roles.
If you’re interested in learning why improving the long-term future might be a key moral priority, read our profile on longtermism.
Contributors
This profile was last updated 20/12/2022. Thanks to Christian Tarsney for helpful feedback on this profile. All errors remain our own. Learn more about how we create our profiles.
Subscribe to the Topic Discovery Digest
Where next?
Our recommended research directions
Explore areas where we think further research could have a particularly positive impact on the world.
Apply for coaching
Want to work on one of our recommended research directions? Apply for coaching to receive personalised guidance.
Prioritisation research
Learn about doing exploratory research on how to do the most good.
Existential and global catastrophic risks
Learn about doing research focused on protecting humanity from large-scale catastrophes.