Eticas

The gold standard in automated bias detection + responsible AI compliance.

Learning by Exposing BadData

Thanks to artificial intelligence, algorithms can be trained by and learn from data. But what an algorithm does depends a lot on how good the data is. Data could be corrupted, out of date, useless or illegal. In this way, bad data plays an important part of all kinds of decision-making processes and outcomes.

In 2018 Eticas started a #BadDataChallenge. It was born to collect the stories of those people who have suffered the social, ethical or economic consequences derived from poor collection, management, storage and processing of their personal data. The objective was to give visibility to a widespread problem and question the established idea about the superiority of data systems over human decisions.

In 2018 we received a number of stories that were concerning and they confirmed to us that the problem was real. Some people ended up paying for the health insurance of another person, others had wrong names of the parents in their ID, which complicated their identification process constantly. However, we did not receive as many stories as we hoped for. Maybe the timing was not right, social media was already fully incorporated in our daily communication but social mobilization campaigns were just building to become a tool of advocacy. And we also tried to do it on our own, exposing the social impacts of BadData by asking our community to share the stories.

After the campaign in 2018, we knew that the potential of exposing BadData in our life was there. We also knew that the problem was far from being solved. We did not give up on the idea, as our mission is to protect people in the technological processes and the increasing datafication of our society. So we decided to re-launch the campaign and apply the lessons learned.

On the 1st of April 2022 we re-launched the BadData campaign. What did we do differently this time?

  • We use the power of social media and interact with the public. We focus on explaining what different types of BadData exist but also provide practical examples that people can relate to and even have an “aha” effect. We have a powerful visual presentation and easy, uncomplicated ways to share the stories.
  • We also asked friends to support us: building alliances and collaborations, working together is a crucial part of the success. Before starting the campaign, we asked like-minded organizations to join us and share the information about the campaign among their communities and social media channels. We also got Maldita.es, a Spanish communication medium, a non-profit foundation dedicated to fact checking on board. Their support with dissemination has been paramount. Thanks to their support and their dissemination efforts, only in the first 3 weeks of the campaign, we received 45 stories only through their platforms.

The BadData campaign will run for another two hot weeks and will be closed on the 1st of July. We have reached 100 stories and the result is overwhelming. There are testimonies from people experiencing BadData in banking: “The director of my company and I had the account in the same branch. I don’t know how it was possible, but one year the bank charged my boss’s real estate tax to my account.” (translation from Spanish).

Numerous stories from phone and telecommunication sectors that waste people´s time: “I received a payment for two iPhones 12 that I never bought. The charge was made by Vodafone without prior notice since the system linked my account to another client who did make the purchase. We ended up in court where the error was proven.” (translation from Spanish).

And even more stories from the public sector and administration“In the civil registry they married my husband to my mother instead of marrying him to me. They were married for several months, we found that out when we went to a notary to make a deed of separation of property(translation from Spanish)

There are many stories from the healthcare sector that are alarming and can endanger the health and well-being of a person: “(Lua) When I signed up for social security as a child, someone at some point decided to change my name to Luz. This has caused me several confusions when collecting results, prescriptions and delays in appointments. Several years ago I managed to correct the mistake, or so I thought until this summer when I had to go to the emergency room in my town in Galicia. They spent 3 hours trying to find me because in the state system I keep appearing as Luz. On other occasions, when I went to pick up a cytology result that the gynecologist had done on me, they spent half an hour looking for my result with no luck. Finally, someone at some point, checking the surnames, told me that I had a Luis with the same surnames as mine, and indeed, the dates matched and they did not have any Luis who had gone to the gynecologist that day.”

And so many more stories from ecommence, travel and tickets, education, insurance, work etc. If you now start to think that BadData might be everywhere, then you are right, it is indeed all around.

We are very happy that we are receiving so many stories, the campaign has been going very well so far. We are also very thankful to the organizations that support our campaign: Maldita.org EncodeJusticeNoledescasito.

But beyond that we are alarmed by how much BadData is there and we refuse to accept it any further. This project not only aims to raise awareness but also takes place at a key moment in the development of AI regulation in the EU, with which an impact is also expected at the level of technological policies and practices.

Please join our efforts. You can still share your BadData experiences with us or share it with your friends, family and colleagues.