3 ms·

The problem with this analysis is that it assumes that all inputs from partisan news sources are providing a consistent set of observations with differing level

by bane 3y ago

The problem with this analysis is that it assumes that all inputs from partisan news sources are providing a consistent set of observations with differing levels of accuracy. This is demonstrably not the case. Partisan news[1] typically acts more like an extremely lossy compression algorithm that throws away information that doesn't already conform to a predetermined pattern. That information is not retained as "low accuracy", it simply doesn't exist in their reporting. Many people still consider these types of sources as fairly reliable today. On occasion (or perhaps disturbingly often) extremely partisan news sources insert (i.e. repeat, amplify) observations that could be thought to have even a negative accuracy (falsehoods) [2]. The difference in conclusions is thus: suppose we had a computing device that could make the NP complete step (#3 in the post) P instead. It would mean that in such a scenario we could rationally analyze partisan news. What I'm saying is that even given such a device, we could not. Given [1] above we would also need a device that could generate all possible observations and accuracies to attempt to fill in the missing information - which requires infinite computation. But [2] makes it even more intractable, where we have to not only compute all possible observations, but also consider all scenarios where there is no observation to report (with it's own set of accuracies) -- an even larger infinity. The article's supposition can be resolved simply by observing enough partisan news (reported from every partisan angle) that you can put error bars on the reported information. Reality is such that you could observe all partisan news and still not know what's going on because there would simply be information gaps that would be irretrievable without infinite computation.

RationalDino 3y ago
I explicitly make no such assumptions. And the fact that I didn't, really matters here. While I assume that you get true (though filtered) statements from some on each side, I'm also assuming that there are unreliable (read dishonest) sources as well. And each side is making the case that the other side can't be trusted. Which therefore means that we should expect (as actually happens) to be getting inconsistent information from different sources. I didn't walk through it in detail, but that is the source of the problem. I start out convinced by one side. Then I observe someone I trust confirming someone I don't trust. Now I wind up trusting someone on the other side more. But as I iterate through my cascade of belief updates, this creates a feedback loop. Strengthening my belief that that person is trustworthy, strengthens my belief in other things that they claimed. Which causes me to trust others on that side more, and trust "my" side less. Which causes me to reevaluate more statements, and so on. The critical question is, "Does this feedback loop die down, or does it continue until I switch sides?" That's the numerically intractable problem which computer scientists have proven to be NP-hard. And being forced to choose between two networks of conflicting beliefs is exactly the situation which causes the biggest problems. Moving on, I agree that being able to evaluate the Bayesian update would not make this computable. I offered a straightforward example - unproven statements of mathematics, encoded as statements about economics. But your current line of argument is exactly the argument of the post that I was responding to.
- bane 3y ago
  I don't really disagree with the concept in the post, other than to say that it's always assuming that there's some information (observations) coming in on any given topic. The nuance I'm arguing for is that truly partisan news is often as much a practice of complete omission than inaccurately reported observations -- something that's....umm....omitted here, thus providing a salient example. This results in an over focus on belief networks and switching beliefs, when the result is more complex. Many people hold no particular beliefs whatsoever on most topics. When pressed they'll generate a "belief proxy" at random until the partisan sources their mental models are fed by gives them an observation to hold to. It's perhaps not mathematical, but it's highly observable -- particularly on topics that are likely to be highly partisan.
  - RationalDino 3y ago
    I'm focused on what we do with information we receive. I have no idea what one is supposed to do about unknown information which was not received.
    - bane 3y ago
      As Donald Rumsfeld famously said when working his way through the permutations in a confusion matrix: "Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don't know we don't know. And if one looks throughout the history of our country and other free countries, it is the latter category that tends to be the difficult ones." Probably more actionable, https://www.amazon.com/Psychology-Intelligence-Analysis-Richards-Heuer/dp/B0016OST3O https://www.amazon.com/Psychology-Intelligence-Analysis-Rich...

by bane 3y ago

RationalDino 3y ago

I explicitly make no such assumptions. And the fact that I didn't, really matters here. While I assume that you get true (though filtered) statements from some on each side, I'm also assuming that there are unreliable (read dishonest) sources as well. And each side is making the case that the other side can't be trusted. Which therefore means that we should expect (as actually happens) to be getting inconsistent information from different sources. I didn't walk through it in detail, but that is the source of the problem. I start out convinced by one side. Then I observe someone I trust confirming someone I don't trust. Now I wind up trusting someone on the other side more. But as I iterate through my cascade of belief updates, this creates a feedback loop. Strengthening my belief that that person is trustworthy, strengthens my belief in other things that they claimed. Which causes me to trust others on that side more, and trust "my" side less. Which causes me to reevaluate more statements, and so on. The critical question is, "Does this feedback loop die down, or does it continue until I switch sides?" That's the numerically intractable problem which computer scientists have proven to be NP-hard. And being forced to choose between two networks of conflicting beliefs is exactly the situation which causes the biggest problems. Moving on, I agree that being able to evaluate the Bayesian update would not make this computable. I offered a straightforward example - unproven statements of mathematics, encoded as statements about economics. But your current line of argument is exactly the argument of the post that I was responding to.

bane 3y ago
I don't really disagree with the concept in the post, other than to say that it's always assuming that there's some information (observations) coming in on any given topic. The nuance I'm arguing for is that truly partisan news is often as much a practice of complete omission than inaccurately reported observations -- something that's....umm....omitted here, thus providing a salient example. This results in an over focus on belief networks and switching beliefs, when the result is more complex. Many people hold no particular beliefs whatsoever on most topics. When pressed they'll generate a "belief proxy" at random until the partisan sources their mental models are fed by gives them an observation to hold to. It's perhaps not mathematical, but it's highly observable -- particularly on topics that are likely to be highly partisan.
- RationalDino 3y ago
  I'm focused on what we do with information we receive. I have no idea what one is supposed to do about unknown information which was not received.
  - bane 3y ago
    As Donald Rumsfeld famously said when working his way through the permutations in a confusion matrix: "Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don't know we don't know. And if one looks throughout the history of our country and other free countries, it is the latter category that tends to be the difficult ones." Probably more actionable, https://www.amazon.com/Psychology-Intelligence-Analysis-Richards-Heuer/dp/B0016OST3O https://www.amazon.com/Psychology-Intelligence-Analysis-Rich...