3 ms·

Technically speaking, if every day you take an uniformly random selection of news articles, on average what you read will not be biased. Or, to be precise, it w

Technically speaking, if every day you take an uniformly random selection of news articles, on average what you read will not be biased. Or, to be precise, it will follow the same biases as the population of news articles you sample from.

denton-scratch 4y ago
> what you read will not be biased. Or to be precise [...] Your "to be precise" sentence states directly that what you read WILL be biased.
strogonoff 4y ago
Replace “news articles” with “facts” and you’re closer to the point, but then how useful would that input be to you?
- ethbr0 4y ago
  The difference is that facts are harder to consume than news articles, as the latter is essentially packaged and distributed collections of the former. Furthermore, the intent of parent's bias averaging, as I understand it, is... 1) There are a finite number of major biases axes in mass media (e.g. political parties, wealthy-poor, conservative-liberal, etc.), ergo averaging these out is a helpful reduction of major biases 2) It assumes the underlying material has a bias, and seeks to counteract it. If you build a fact-based method, you leave yourself open to treating biased opinions as facts, if they're sufficiently masked
  - strogonoff 4y ago
    Indeed, the first challenge to unbiased approach would be the impossible in practical terms task of separating facts from fiction. I don’t know if it is clear but I really meant it that unbiased news is not a possibility due to fundamental limitations of information theory and finiteness of individual information processing capabilities; self-awareness and bias transparency is the only way. > helpful reduction of major biases Only if those major biases are opposite between media. Imagine biases as vectors pointing in arbitrary directions. Best case is when bias vector in one cluster is compensated by exactly opposite bias vector in another cluster, but how often do you expect that to happen? A more typical scenario is a vector pointing in merely a different, rather than opposite, direction. A worst-case scenario, and also common (because bias manifests itself in selection if facts to report, and most sources report on the same events), is when all bias vectors point in a similar direction.
    - ethbr0 4y ago
      In your original (upthread) argument that "unbiased news is not a possibility", are you stating that as a fundamental consequence of the news' lossy compression of {all facts} into a smaller product... or something else? It's unclear. As for averaging out, yes, the vector model was what I was using internally. Disclosure: I'm US-based, so our politics are traditionally bipolar as a result of our election system (in contrast to Europe et al.). I would be surprised if vectors don't cancel often though, because think what we're really talking about in a capitalist information space: popular perspectives that are profitably marketable. What is the easiest way to find an audience? Do what your competitors are not. So on the whole, high revenue and production cost media (tv, web) in a capitalist information space will converge into bipolar pairs on any given issue, in order to maximize market share. In lower cost media or operations, it will more likely be a random scattering of vectors.
      - ryanbrunner 4y ago
        > Disclosure: I'm US-based, so our politics are traditionally bipolar as a result of our election system (in contrast to Europe et al.). There's a whole lot of bias that's "baked in" with US politics / media. There's broad consensus on neoliberal capitalism, personal freedoms vs collectivism, common ground on opposing states like Russia and China. On a few wedge issues you will average out to a common ground, but ultimately you're not eliminating bias, you're just eliminating a relatively small bias between left-wing and right-wing US and replacing it with an "overall US" bias.
        ethbr0 4y ago
        From mass market sources, true. Probably because they're aiming towards the common American, who generally has similar feelings. Ergo, they're that + whichever politically-polarized biases. As for whether averaging that out is useful...? I'd be happy to have news from a "overall US" bias, without the right/left political spin.
ryanbrunner 4y ago
It's almost impossible to arrive at a pure random sampling like that. As an easy example, if you limit yourself to English news articles, you're going to have a pretty large bias towards "the West", which right now probably includes a strong bias against China or Russian's viewpoints on things (not to say that including those viewpoints would be more truthful necessarily, but certainly less biased).

denton-scratch 4y ago

> what you read will not be biased. Or to be precise [...] Your "to be precise" sentence states directly that what you read WILL be biased.

strogonoff 4y ago

Replace “news articles” with “facts” and you’re closer to the point, but then how useful would that input be to you?

ethbr0 4y ago
The difference is that facts are harder to consume than news articles, as the latter is essentially packaged and distributed collections of the former. Furthermore, the intent of parent's bias averaging, as I understand it, is... 1) There are a finite number of major biases axes in mass media (e.g. political parties, wealthy-poor, conservative-liberal, etc.), ergo averaging these out is a helpful reduction of major biases 2) It assumes the underlying material has a bias, and seeks to counteract it. If you build a fact-based method, you leave yourself open to treating biased opinions as facts, if they're sufficiently masked
- strogonoff 4y ago
  Indeed, the first challenge to unbiased approach would be the impossible in practical terms task of separating facts from fiction. I don’t know if it is clear but I really meant it that unbiased news is not a possibility due to fundamental limitations of information theory and finiteness of individual information processing capabilities; self-awareness and bias transparency is the only way. > helpful reduction of major biases Only if those major biases are opposite between media. Imagine biases as vectors pointing in arbitrary directions. Best case is when bias vector in one cluster is compensated by exactly opposite bias vector in another cluster, but how often do you expect that to happen? A more typical scenario is a vector pointing in merely a different, rather than opposite, direction. A worst-case scenario, and also common (because bias manifests itself in selection if facts to report, and most sources report on the same events), is when all bias vectors point in a similar direction.
  - ethbr0 4y ago
    In your original (upthread) argument that "unbiased news is not a possibility", are you stating that as a fundamental consequence of the news' lossy compression of {all facts} into a smaller product... or something else? It's unclear. As for averaging out, yes, the vector model was what I was using internally. Disclosure: I'm US-based, so our politics are traditionally bipolar as a result of our election system (in contrast to Europe et al.). I would be surprised if vectors don't cancel often though, because think what we're really talking about in a capitalist information space: popular perspectives that are profitably marketable. What is the easiest way to find an audience? Do what your competitors are not. So on the whole, high revenue and production cost media (tv, web) in a capitalist information space will converge into bipolar pairs on any given issue, in order to maximize market share. In lower cost media or operations, it will more likely be a random scattering of vectors.
    - ryanbrunner 4y ago
      > Disclosure: I'm US-based, so our politics are traditionally bipolar as a result of our election system (in contrast to Europe et al.). There's a whole lot of bias that's "baked in" with US politics / media. There's broad consensus on neoliberal capitalism, personal freedoms vs collectivism, common ground on opposing states like Russia and China. On a few wedge issues you will average out to a common ground, but ultimately you're not eliminating bias, you're just eliminating a relatively small bias between left-wing and right-wing US and replacing it with an "overall US" bias.
      - ethbr0 4y ago
        From mass market sources, true. Probably because they're aiming towards the common American, who generally has similar feelings. Ergo, they're that + whichever politically-polarized biases. As for whether averaging that out is useful...? I'd be happy to have news from a "overall US" bias, without the right/left political spin.

ryanbrunner 4y ago

It's almost impossible to arrive at a pure random sampling like that. As an easy example, if you limit yourself to English news articles, you're going to have a pretty large bias towards "the West", which right now probably includes a strong bias against China or Russian's viewpoints on things (not to say that including those viewpoints would be more truthful necessarily, but certainly less biased).