Quotes from Kate Crawford

We should always be suspicious when machine-learning systems are described as free from bias if it's been trained on human-generated data. Our biases are built into that training data.

If we start to use social media data sets to take the pulse of a nation or understand a crisis - or actually use it to deploy resources - we are getting a skewed picture of what is happening.

It is a failure of imagination and methodology to claim that it is necessary to experiment on millions of people without their consent in order to produce good data science.

Only by developing a deeper understanding of AI systems as they act in the world can we ensure that this new infrastructure never turns toxic.

Many of us now expect our online activities to be recorded and analyzed, but we assume the physical spaces we inhabit are different. The data broker industry doesn't see it that way. To them, even the act of walking down the street is a legitimate data set to be captured, catalogued, and exploited.

Surveillant anxiety is always a conjoined twin: The anxiety of those surveilled is deeply connected to the anxiety of the surveillers. But the anxiety of the surveillers is generally hard to see; it's hidden in classified documents and delivered in highly coded languages in front of Senate committees.

Biases and blind spots exist in big data as much as they do in individual perceptions and experiences. Yet there is a problematic belief that bigger data is always better data and that correlation is as good as causation.

Hidden biases in both the collection and analysis stages present considerable risks and are as important to the big-data equation as the numbers themselves.

Vivametrica isn't the only company vying for control of the fitness data space. There is considerable power in becoming the default standard-setter for health metrics. Any company that becomes the go-to data analysis group for brands like Fitbit and Jawbone stands to make a lot of money.

Data and data sets are not objective; they are creations of human design. We give numbers their voice, draw inferences from them, and define their meaning through our interpretations.

The amount of money and industrial energy that has been put into accelerating AI code has meant that there hasn't been as much energy put into thinking about social, economic, ethical frameworks for these systems. We think there's a very urgent need for this to happen faster.

Rather than assuming Terms of Service are equivalent to informed consent, platforms should offer opt-in settings where users can choose to join experimental panels. If they don't opt in, they aren't forced to participate.

When dealing with data, scientists have often struggled to account for the risks and harms using it might inflict. One primary concern has been privacy - the disclosure of sensitive data about individuals, either directly to the public or indirectly from anonymised data sets through computational processes of re-identification.

We should have equivalent due-process protections for algorithmic decisions as for human decisions.

We need to be vigilant about how we design and train these machine-learning systems, or we will see ingrained forms of bias built into the artificial intelligence of the future.

As we move into an era in which personal devices are seen as proxies for public needs, we run the risk that already-existing inequities will be further entrenched. Thus, with every big data set, we need to ask which people are excluded. Which places are less visible? What happens if you live in the shadow of big data sets?