Data polluted with patterns of human thinking can only be dangerous

“You could use data to promote social equity, to fight hidden institutional biases, policies and practices,” says Eline Chivot, Senior Policy Analyst at the Center for Data Innovation. At the same time, collection of too much data could lead to a surveillance state. But what is too much?

“Hardly anything runs without data. We see the rise of data driven-companies and platforms. It seems that data is the blood in a government’s or company’s body - it makes it more agile, more resilient, open, and entrepreneurial,” Ramojus Reimeris, Head of Unit at STRATA, said in a panel discussion about data-driven decision making during the GovTech Week_.

According to Reimeris, we are collecting vast amounts of data without even knowing how it might be used further on. “Where’s the red line?” he asked the panelists.

Katarzyna Szymielewicz, human rights activist, lawyer, and president of the Panoptykon Foundation, argued that data collection always comes with negative externalities.

“For something that is socially questionable, potentially harmful, and even flawed in essence, I would say any data is too much data,” she said.

Why do we need so much data?

“Data leads to better understanding and decision making among individuals, businesses, governments, and government agencies that are using data to cut costs, prioritize social services, and keep citizens safe. They can use data to deliver better services to citizens,” said Eline Chivot.

According to her, there are a lot of positive outcomes in data collection: it helps promote social equity, to fight hidden institutional biases, policies, and practices.

However, government institutions are quite slow to adopt new technologies.

“The public sector, in particular, lags behind when it comes to harnessing the potential of data and transitioning towards digital for several reasons. They are outsourcing digital, or there’s a lack of software culture, or a risk of civil servants not trusting the work of technologies,” elaborated Eline Chivot.

To this day, algorithms work in a way that the more data they collect, the more accurate the technology is likely to be. Also, the quality of the data is extremely important. 

“We also have to make sure that it is properly cleaned, shared, collected in order to really find applications that are accurate and beneficial,” said Eline Chivot.

Irrelevant, or confusing, or polluted with certain patterns of human thinking, data can only be dangerous, and collecting more of it will not help us solve the problem,

said Katarzyna Szymielewicz.

Katarzyna Szymielewicz from the Panoptykon Foundation reckons that feeding data about human behavior to an algorithm might not produce any accurate results.

“It is quite tricky to collect accurate data about human behavior. It is easier to do that with cars, crops, and weather,” she said.

It’s impossible to predict the behavioral patterns of individuals, no matter what kind of data you have.

“The best you can achieve is to repeat certain stereotypes. Irrelevant, or confusing, or polluted with certain patterns of human thinking, data can only be dangerous, and collecting more of it will not help us solve the problem,” said Katarzyna Szymielewicz.

When have we collected enough?

“I don’t have an answer to when it becomes too much. It’s very contextual,” said Katarzyna Szymielewicz when asked how much data is too much.

To her, it depends on the purpose of data collection - are you trying to cure cancer, doing research on people’s behavior to develop better psychological treatments, or just need data to increase sales?

“For something that is socially questionable, potentially harmful, and even flawed in essence, I would say any data is too much data,” she said.

Katarzyna Szymielewicz gave an example of an online marketing industry that is heavily data-driven.

“It is not socially positive, does not reward people, and is even questioned by the industry itself. There’s a lot of research showing how flawed the whole behavioral advertising models are. Needless to say, they can be harmful to people,” she said.

Also, sometimes it’s hard to say whether statistical non-personal data is in fact non-personal.

“To what extent data generated by cars is at the same time data of their users, data certainly describes the way the machine is used, and the places where the machine goes. It is revealing quite a lot of personal information, even if at the beginning we could think that we are just collecting the data about the engine of a car,” she said.

Data collection during the pandemic

Countries worldwide have been creating so-called virus tracing apps. They were, for example, launched in France, Finland, the UK, and other countries. In western countries, most of them are voluntary, so only people who consent to give their personal information to the governmental institutions install the app. That is also the reason why some of them are extremely unpopular.

According to Eline Chivot, European data protection rules limit the processing of personal data but they do allow for public health exceptions in the context of pandemics with certain limits and safeguards. 

“You see that in Finland, the COVID app seems to be working rather well - 20% of the population downloaded the app within the first day. But in France it’s only 70 people,” she said.

She believes that there has to be public trust for this to work: “Citizens are willing and ready to make concessions to some of their liberties and freedoms in these circumstances. But some worry that these datasets could be put for other use in the future.”

Meanwhile, Katarzyna Szymielewicz concluded that COVID-19 proved to be very useful in terms of data collection.

“In Europe, we moved to the discussion from “let’s collect it all, let’s look at personal data and employ every technology we have to observe human behavior,” to solutions that are very privacy-friendly, basically using only anonymous data. I refer now to the standard that was finally adopted by Google and Apple for a COVID-19 tracking app. The appetite for collecting more data and storing it for the future has been curbed. At the end of the day, we have solutions that do not expose people’s privacy,” she said.

You can see the full discussion here. (Starts at 2:18)