rfl-urbaniak
commented
4 months ago We ned to come with a realistic structure for the book. I still think the book should have these three parts:
How evidence is evaluated
How decisions are made
Simulations about truth and fairnessR: I'm skeptical about the feasiblity of the last part. (1) literature on fairness is huge, we haven't even touched it properly. (2) time constraints.
We have been working on a few things:
higher order uncertainty,
coherence and stories (Rafal's existing work)
awareness growth.
These can be our three themes in the evidence evaluation. The point here is to show that legal reasoning with evidence is more complex than just Bayesian updating and we sketch a framework to capture this greater complexity.R: also: missing evidence, we have an example and a BN elaboration of that somewhere, right?
(Perhaps awareness growth is just another source of higher order uncertainty?)
R: Interesting, will ponder over this.
We would need to have three compelling examples of legal reasoning with evidence where these three dimensions are used. We already have an example for higher-order uncertainty, debates about DNA evidence databases etc. For stories and coherence, I think we might want to draw from Old Chief case.
R: one of my papers with Alicja has extensive coherence calculations for the Sally Clark case, and a bunch of other examples (some involving murders) we might use those and perhaps one more realistic example. Have you looked at these?
We need to show what types of reasoning pattern does story/coherence allows us to capture.
R: Sure, a point of departure is to take a new look at the two existing papers.
Finally, I think awareness growth has to do with cross-examination and the openness of the possibility space.
We might need a chapter that takes a step back and reflects on the role of formal methods for evidence evaluation, what are they good at and what are they bad. What are their limits and advantages? This chapter would frame our contribution.
So I see four chapters in total for this section.
PART 2: Decision making We have not worked much on this section. We have a chapter on the conjunction paradoxes. This is about decision theory. We need to discuss what we think about this. Perhaps our decision theory is just multidimensional -- consider the guilt probability, the coherence of the story, higher order uncertainty and resistance to objections, weigh all these dimensions and then make a decision. Is this the theory we have in mind? It does not need to be formal. Perhaps we could just explore possible decision rules based on these difference dimensions.
Suggestion: the decision theory is still a probability threshold but the preliminary requirements of story coherence, etc. must be met. The problem is that we need a threshold for these others requirements: how coherent should the story be, high low the higher order uncertainty, etc.?
Or perhaps it would be enough to focus on the limitation of just an expected utility maximization approach. Since, as we show in the earlier section, there are so many other dimensions of uncertainty, a narrow focus on expected utility maximization of guilt probability is misleading.
R: I think we should avoid getting into elaborate decision-theoretic arguments. I think the minimal viable version has it that in line with some of my/our/existing earlier papers the existing probabilistic explications are flawed, and that standards formulated in terms of HOPs and narrations do better.
PART 3: Simulation about truth and fairness. We have not talked about this much, but I think it is important because it illustrates the value of part 1 and part 2.
R: see above
We ned to come with a realistic structure for the book. I still think the book should have these three parts:
I think the three parts go together.
PART 1: How evidence is evaluated.
We have been working on a few things:
(Perhaps awareness growth is just another source of higher order uncertainty?)
We would need to have three compelling examples of legal reasoning with evidence where these three dimensions are used. We already have an example for higher-order uncertainty, debates about DNA evidence databases etc. For stories and coherence, I think we might want to draw from Old Chief case. We need to show what types of reasoning pattern does story/coherence allows us to capture. Finally, I think awareness growth has to do with cross-examination and the openness of the possibility space.
We might need a chapter that takes a step back and reflects on the role of formal methods for evidence evaluation, what are they good at and what are they bad. What are their limits and advantages? This chapter would frame our contribution.
So I see four chapters in total for this section.
PART 2: Decision making We have not worked much on this section. We have a chapter on the conjunction paradoxes. This is about decision theory. We need to discuss what we think about this. Perhaps our decision theory is just multidimensional -- consider the guilt probability, the coherence of the story, higher order uncertainty and resistance to objections, weigh all these dimensions and then make a decision. Is this the theory we have in mind? It does not need to be formal. Perhaps we could just explore possible decision rules based on these difference dimensions.
Suggestion: the decision theory is still a probability threshold but the preliminary requirements of story coherence, etc. must be met. The problem is that we need a threshold for these others requirements: how coherent should the story be, high low the higher order uncertainty, etc.?
Or perhaps it would be enough to focus on the limitation of just an expected utility maximization approach. Since, as we show in the earlier section, there are so many other dimensions of uncertainty, a narrow focus on expected utility maximization of guilt probability is misleading.
PART 3: Simulation about truth and fairness. We have not talked about this much, but I think it is important because it illustrates the value of part 1 and part 2.