Action-Impact Tradeoffs

coming soon…

  1. Predictions Actions Redux - An attempt to synthesize a complexity axis for decision making leads me back to the interplay between predictions and actions.

  2. Theories of recourse - A proposed parameterization of theories with recourse in terms of rates of action and impacts of the actions.

  3. You Take Away The Context - The greedy algorithm for contextual optimization highlights how more information should make decision problems less uncertain.

  4. Bracing for Impact - Revisiting 2018's tour of reinforcement learning, but now with a focus on tradeoffs between modeling, prediction, and action.

  5. All models are wrong, but some are dangerous - Sometimes better predictions lead to worse outcomes. I explain by revisiting a 50-year-old paradox in control theory.

  6. An argmin mailbag - Engaging with some of the responses to this series.

  7. The Soothing Warmth of Vacuum Tubes - An introduction to the feedback amplifier and what it can teach us about uncertainty, predictions, and decisions.

  8. The same but different - More examples of when good predictions are insufficient for good decisions. There are probably too many equations in this post, but at least there are no Laplace transforms.

  9. Lamenting the Laplace transform - Biting the bullet and blogging about the frequency domain. One of the biggest technological miracles of the 20th century is that these ornate calculations have any connection to actual practice.

  10. Complex Sensitivity Analysis - The upside of working with the frequency domain is that it simplifies analysis of uncertainty and robustness. And it helps to highlight the sweet spots of uncertainty quantification.

  11. Respect the Unstable - Gunter Stein’s Bode Lecture “Respect the Unstable” reminds us that decision systems have limits, and ignoring them leads to disaster. Had we already hit these limits in 1989?

  12. Purpose Driven Uncertainty Quantification - Takeaways from a month of blogging about the interplay between uncertainty quantification, feedback, and decision making.