This draft is just a sketch that I hope to continue at some point. Not complete, not finished, mostly jotting down some thoughts. I think there's a lot to do in this area, and I'd love it if there'd be something like a roadmap.

Modeling with uncertainty

1. Are distributional CEAs more accurate than point estimates? If so, by how much?
2. What are the best practices for using results from the literature? They are usually expressed as the expected average effect size and a frequentist confidence interval, and we seek a subjective credence distribution.
3. Choice of distribution - when should one express their subjective estimate as a Normal / Lognormal / Beta / Other? How much do these choices matter?
4. How do we best take model error into account?

As input to decision-making

How do we best take the resulting CE into account in decision-making?

1. How to account for and present different types and sources of uncertainty?
2. How to allow for different types of risk preferences? 
3. Given a risk-aversion preference, how should we build a portfolio of donations/projects?

Uncertainty Quantification with Experts

https://osf.io/vk3bf/ Uncertainty quantification with experts: present status and research needs (2021)

Prior knowledge elicitation: The past, present, and future [review paper 2023] 

GPI Research Agenda (draft)

The following is part of the work-in-progress economics research agenda^[1] by the Global Priorities Institute, section 1.3 on "Conceptual issues in cost-effectiveness analysis", where they list four questions.  

What is the distribution of cost-effectiveness?

The distribution of cost-effectiveness within and between charitable causes may have important implications for priority setting. For instance, it could provide a Bayesian prior for adjusting noisy cost-effectiveness estimates of interventions that have not yet been thoroughly evaluated. Empirically, what is the distribution of cost-effectiveness within and between the most impactful charitable causes, and what does this imply for priority setting? (Jamison et al. 2006; Vivalt 2015; 2020)

Accounting for externalities

Many interventions in global health and beyond have significant externalities on individuals other than the direct recipients of the intervention. How should such indirect effects be incorporated in cost-effectiveness analysis?

Comparing extending/creating life with improving life

How do social welfare gains from improvements in people’s quality of life (e.g. improvements in health or consumption) compare to social welfare gains from extending people’s lives or bringing new people into existence? (Cf. Luyten et al. 2022) 

Metrics for health-related interventions - DALYs/WELLBYs

The cost-effectiveness of health interventions is often measured in terms of disability-adjusted life years (DALYs) per dollar. However, the DALY approach has well-known limitations, and there are other approaches such as using wellbeing-adjusted life years (WELLBYs) (e.g., Layard and Oparina 2021). How do these metrics differ in evaluations of cost-effectiveness of different interventions, and which metrics are most appropriate?

1. ^{^}

   Link here, accessed 19.11.23. More updated version may appear on this page. The list is in bullet points, I've added the titles in bold for clarity.