Instructor Notes

The article breaks down the process of generating forecasts for management into 4 phases. Briefly define each phase.

  • Scoping
  • Development
  • Delivery
  • Evaluation

What are the ethical issues that can emerge in the scoping phase?

  • Conflicts of interest
    • Scientic motivations vs. user group needs
    • Interests of stakeholders losing out to the technical and scientific challenges of making the forecast
    • Selection of “who” the stakeholders are can lead to advantaging already advantaged groups (e.g., ‘industry winners’ over other groups).
    • Conflicts between different stakeholders
  • Ecosystem Health
    • Unintended consequences - Could forecasts for fish result in overfishing by allowing more efficient targeting of the fisher?
    • Making forecasts that can impact the extraction of a species without working with management to offset those negative consequences could be a problem for some types of forecasts.

What ethical issues can emerge in the development phase?

  • Related to technical judgements about the system and communicating them to the end-user
  • Skill assessment
    • Need unbiased assessments about how well a forecast is performing and what contexts it can be used in
    • If skill can’t be evaluated discussion of potential risks with stake holders is critical
    • Use best practices for evaluating forecasts
  • Conveying Uncertainty
    • Uncertainty needs to be conveyed to end users so they can understand the probability of a certain outcome.
    • Can be particularly important when presenting in development forecasts (also: “For Research Only” labels)
    • May be context dependent (e.g., season of forecast)
    • Can be communicated both quantitatively & qualitatively

What are the ethical issues that emerge in the delivery phase?

  • Issues related to maintaining the delivery of products and communication/feedback with stakeholders.
  • Delivery of products
    • Research endeavors have a limited time scope. Ongoing delivery of products beyond this time scope should be discussed with end-users upfront.
      • Is there a plan for making these forecast self-sustaining?
      • A partner willing to take on the task of running them and communicating results?
      • Or will the project die after a period of time?
      • “Ultimate solution is to pass the forecast system to an operational system”
    • Should forecasts be made available prior to peer review?
      • Should forecasts only be made to end users after scientific peer-review?
      • Does the risk of making bad forecasts cause issues for the eld of forecasting in general, especially if these forecasts have not been properly communicating their research-level nature or the uncertainty involved?
      • Interesting tensions between the speed of feedback from end users and the slowness of the scientific evaluation process.
    • How to best communicate forecasts to end users depends on end user needs (e.g., web vs. text messages)
  • Education
    • Need to work with end users to improve their capacity to interpret information
    • Different users in different places on the simplicity-details tradeoff
  • Delivery failure: Sometimes delivery failure is the ethical decision if issues with the models are uncovered.
  • Equity for users
    • Delivery of forecasts may advantage/disadvantage different groups
    • Who is receiving the forecasts? Who isn’t?
    • Data from end users or humans? Might be confidentiality issues re: data
  • Unintended consequences
    • Media coverage and public discussion of a forecast intended for a specific group
    • Higher harvest efficiency can mean lower wages for workers who spend less time working
    • Faster understanding of the system - with positive and negative policy outcomes

What are the ethical issues that emerge in the evaluation phase?

  • Issues arising from the performance of the forecast - did it achieve its overall goal (e.g., improved decision-making or conservation)?
  • Review of performance
    • Should forecasts continue? If continuing the forecast works against the overall goal, then it may be necessary to ethically stop forecasting. Should you continue forecasts that help people better find and exploit a limited resource if the management infrastructure is not I’m place to deal with that?
    • Is the forecast causing active harm to end users, especially if that harm is due to limits on the forecast itself?

What are the main principles suggested by the authors for ethical forecasting?

  1. Open and transparent
  2. Avoid making forecasts that have unregulated impacts
  3. Use best practices for assessing skill
  4. Don’t ignore uncertainty
  5. Plan for stakeholder expectations re: continued delivery
  6. Engage in training and communication to help stakeholder better interpret results
  7. Explore the sensitivity of forecasts to missing data products
  8. Be prepared to pull the plug if quality is compromised
  9. Be vigilant for inequity
  10. And for unintended consequences
  11. Always keep the goal of your forecast in mind