"Indeed, when I asked the company whether it would permit an external audit of its News Feed workflow and algorithms to prove that there are no hidden or inadvertent biases against stories critical of itself, a company spokesperson repeated its statement that it believed there were no biases, but did not respond to two separate requests asking whether it would permit an external audit to prove it.
...
Machine learning approaches are especially troubling, as the company continues to refuse to release any information about the functioning and accuracy of its models, even as they play an ever-greater role in shaping what two billion people can see and talk about in its walled garden.
Most recently, when asked about its efforts to train machine learning models to autonomously decide what is "fake news," the company responded that it was using a large number of signals (though it declined to elaborate on the full list of signals used) to train computerized models to fully autonomously scan what is being posted and discussed on Facebook and identify new stories the algorithms believe are false - all without any human intervention.
...
Despite controlling what nearly a quarter of the earth's population sees and says online in its walled garden, the company has survived nearly a decade and a half of privacy outcries without ever having to open up and give its users even the slightest insight into how they are being manipulated, moderated and commercialized.
...
Putting this all together, Facebook's utopian vision has devolved into a surveillance dystopia in which even its programmer creators can't be certain how or why it makes the decisions it does.
In the end, the telescreen's of Orwell's 1984 only surveilled the citizenry at random, while Facebook's unblinking algorithms never let us out of their sight, silently shaping what we are able to see and say without us having any right to understand the rules they quietly enforce, while even their engineer creators are not fully aware of the ramifications of the myriad inadvertent decisions that went into their programming."
...
Machine learning approaches are especially troubling, as the company continues to refuse to release any information about the functioning and accuracy of its models, even as they play an ever-greater role in shaping what two billion people can see and talk about in its walled garden.
Most recently, when asked about its efforts to train machine learning models to autonomously decide what is "fake news," the company responded that it was using a large number of signals (though it declined to elaborate on the full list of signals used) to train computerized models to fully autonomously scan what is being posted and discussed on Facebook and identify new stories the algorithms believe are false - all without any human intervention.
...
Despite controlling what nearly a quarter of the earth's population sees and says online in its walled garden, the company has survived nearly a decade and a half of privacy outcries without ever having to open up and give its users even the slightest insight into how they are being manipulated, moderated and commercialized.
...
Putting this all together, Facebook's utopian vision has devolved into a surveillance dystopia in which even its programmer creators can't be certain how or why it makes the decisions it does.
In the end, the telescreen's of Orwell's 1984 only surveilled the citizenry at random, while Facebook's unblinking algorithms never let us out of their sight, silently shaping what we are able to see and say without us having any right to understand the rules they quietly enforce, while even their engineer creators are not fully aware of the ramifications of the myriad inadvertent decisions that went into their programming."
Source:
https://www.forbes.com/sites/kalevleetaru/2018/04/02/faceboo...