intro chapter with objections draft

[marcellodibello]

added first rough draft of intro chapter with objections

[marcellodibello]

started drafting chapter with objection to legal probabilism, making good progress, added one objection: where do the number comes from? will continue working in the following days on other objections.

[marcellodibello]

more progress on intro chapter with objections to legal priobabilism. started to work on cases that illustrate why a high probability cannot be the sole measure of uncertainty for guiding trial decision making

[marcellodibello]

interesting new find, this paper applies chain event graphs (CEG) to the modeling of evidence and activity-level propositions in a case of drug and money laundering. "Chain event graphs for assessing activity-level propositions in forensic science in relation to drug traces on banknotes:" by Gail Robertson, Amy L. Wilson2 and Jim Q. Smith

full paper here: https://arxiv.org/pdf/2404.02778v1.pdf

(for the general theory of CEG, see paper in Artificial Intelligence: see https://www.sciencedirect.com/science/article/pii/S0004370210000810)

CEGs are modification of Causal Bayesian Networks. Nodes represent events that are temporarily ordered. Each edge is associated with a transition probability. CEGs are suited to represent competing stories or explanations of the evidence. The authors write (p. 3):

"We argue that for cases involving activity-level propositions a staged tree and its associated CEG have many advantages. The storylines proposed by the prosecution and defence can be directly extracted from a CEG for use by a barrister to describe the unfolding of events in the case, which is compelling. The algorithms used in the CEG do not need the delicate construction of random variables to appropriately define any BN representation. Finally, the temporal coherence of represented storylines makes it easier to encourage jurors to implicitly introduce probabilities of events in a logical way, as these are described under the competing propositions."

The problem "where do the numbers come from?" recurs. The authors write (p. 13):

"Given the large number of scenarios considered in the CEG, even if it were possible to carry out the experiments listed above there would still be insufficient data to inform all of the edge probabilities using data alone. Therefore the bulk of the edges in the CEG must be set using expert judgement. This can be combined with sensitivity analysis to see the effect that different choices have on the results."

QUESTION: do you know about CEG @rfl-urbaniak and @Niklewa? should we include it when we talk about narratives and stories and coherence?

[marcellodibello]

cleaned up a bit of the chapter on objections to legal probabilism. it can now be reviewed up to section 6.

four majors objections are outlined:

1. where do the numbers come from?
2. beyond high probability (other dimensions matter: resilience, high order uncertainty, weight, specificity, coherence, etc.)
3. how evidence and hypothesis aggregation works? what is the right modeling tools? what about arguments or stories?
4. beyond Bayesian updating (sometimes we need to reason about the probability model itself or the space of possibilities)

feel free to review up to section 6 @rfl-urbaniak and @Niklewa

you can find each objection in short version under "the challenge in a nutshell"