Epistemic Momentum is code name for my MS Thesis that states:
Adversarial attacks seeking to specifically undermine foundational trust within human- machine teams raise concerns of stability and safety for learning and inference systems in military, healthcare, finance, and other domains.
________ ____ ____
|_ __ ||_ \ / _|
| |_ \_| | \/ |
| _| _ | |\ /| |
_| |__/ | _| |_\/_| |_
|________||_____||_____|
We come to evaluate the evidence we have for a particular belief in many different ways, even in very different cases where our eventual position is the same. In light of sure and corresponding initial evidence, we might come to strong belief in a proposition very quickly. On the other hand, we might begin with complete ambiguity with respect to that same proposition, and yet come to the same conclusion with the later introduction of clear evidence. What can we say about the difference between these two cases?
In particular, if clear evidence follows initial ambiguity, then how does our re-evaluation of that initial set of evidence affect our belief state? How does this compare with having initially clear and strong evidence supporting that same belief? Does this difference tell us something about each of these beliefs and perhaps how we might interpret the next piece of evidence of these positions? This paper puts forth a model of coming to belief where the re-evaluation of prior ambiguous claims creates an epistemic momentum in agents who acquire clear evidence later. This energy within the evaluation of belief evidence provides a how-possibly explanation of the differences in how we come to belief and characterizes a type of polarization as being a natural outcome of conclusions led by a high amount of ambiguity of initial evidence.
Currently file per experiment.
From this directory, python|3 baseline.py will write to output directory. output/results.txt is the matrix of numbers that should generate the correct curve.
3.1g = baseline_2vars 3.1h = baseline_2vars_strong