I believe that many people, including me, put a lot of trust in technology. And why shouldn’t we? Technology in agriculture such as fertilizers and machinery have given us the ability to produce more food than we can eat. We live longer, healthier lives because of technological improvements in the medical field. We can even build giant cities in deserts.
But what about information technologies. Can we trust (future) information-based technologies to solve all our problems? I want to illustrate this with an example.
If you want to know more about your ancestry, or you might want to know if you’re genetically more prone to certain health issues, there’s 23andMe – a FDA-cleared DNA genetics testing start-up firm. For around 200 USD, they can tell you everything you want to know about your health, traits, and ancestry, which is all based on your genes (Freakonomics Radio, 2019).
23andMe is a DNA Genetics Testing start-up in Mountain view, CA.
Well actually, your personalized report is based on the 23andMe DNA database. They match your DNA with everyone who has ever sent in their DNA sample. This database is currently filled with more than 10 million unique people (TheDNAgeek, 2019), supposedly adding to greater result-accuracy. However, the problem is that this database is predominately European-centric. In fact, 80% of all users are from European descent, implying low genetic diversity (Freakonomics, 2019). This is not necessarily a problem if you are white, but it does seem to discourage or even lower accessibility if you’re a person of color because you might wonder if your personalized 23andMe results are less precise.
This is not necessarily the fault of 23andMe, because the company relies on the DNA input of its users. But in order to increase accessibility for everyone, databases should be representative of all populations. Not only in the case of genetic testing, but in all parts of our society in which we use information-based technologies to solve problems. From applications that use face recognition to machine learning models that help determine what advertisements we get to see.
Technology is made by humans. Their ideas, values, and views are expressed in their work, which have serious impact on the people using these technologies (Blauw, 2019). How do we ensure that the people who create our solutions, do so in an unbiased way? I’d like to know what you guys think. Thanks.
I was inspired to write this blog by Sanne Blauw, a reporter at the Correspondent. She wrote an interesting article about the people behind artificial intelligence. If you’re interested, I’ve included a link below. She also made a podcast, if you’re more into that.
Sanne’s article (De Correspondent): decorrespondent.nl/10589/kunstmatige-intelligentie-is-menselijker-dan-het-lijkt-wie-zitten-erachter/154098688546-984c0d3d
Sanne’s (15 min) podcasts: open.spotify.com/episode/2dnsWAa4IxvKqDS1YfYdm4?si=-_lNaYUOTliLC9XSswUwdA
References
Freakonomics Radio. (2019). 23andMe (and You, and Everyone Else). Stephen, J. Dubner/Freakonomics Radio. Retrieved from: https://medium.com/s/freakonomicsradio/23andme-and-you-and-everyone-else-18ab5941c95d
TheDNAgeek. (2019). 23andMe Has More Than 10 Million Customers. Retrieved from: https://thednageek.com/23andme-has-more-than-10-million-customers/
Blauw, S. (2019). Kunstmatige intelligentie is menselijker dan het lijkt. Wie zitten erachter? De Correspondent. Retrieved from: decorrespondent.nl/10589/kunstmatige-intelligentie-is-menselijker-dan-het-lijkt-wie-zitten-erachter/154098688546-984c0d3d
Hey Leon, thank you for sharing such an interesting topic and to show such an innovative start-up. I tend to agree with you when you state that you put a lot of trust in technology. In fact, I do too because I think that in a foreseeable future we will no choice but to rely on technology for almost everything in our lives.
It is really interesting to know that there are possibilities to know everything there is to know regarding our ancestry, genes, traits and health. And it is fascinating to know that this is only possible due to information-based technologies. The fact, that this startup is able to match our DNA with other samples is amazing. However, I must say that I feel quite concerned regarding this due to privacy reasons and data scandals, as it relies on data input of its users. Furthermore, I wonder to what extent is this technology reliable, too. AI-based models often rely on millions of pieces of data and sometimes there such a complex underlying structure that diagnostics can be hard to interpret. In addition to this, the fact that these systems might have hidden biases derived from data (for instance, in this case the fact that 80% of all users are European), the insights generated from it are far from being 100% reliable which puts in risk consumers that pay 200 to obtain this kind of information. In fact, these AI systems often produce statistical truths rather than in-depth analysis (with verifiability) which makes it difficult to prove that they can work in all cases. On top of this, the fact that the people behind this technology certainly express their values and ideas when creating this system will, for sure, contribute to an increasing bias in the results it generates.
Hey Pedro, thanks for your message.
I want to respond to your two main concerns: 1) privacy and hacks, and 2) reliability of results.
Anna Wojcicki – CEO of 23andMe – has addressed concerns on privacy and data scandals in an interview with Freakonomics Radio. Since they deal with health information, they abide to the HIPAA law regulation. This means that they treat customer health information very carefully and nothing is shared to 3rd parties if you don’t want to. In fact, if you do want that, you can opt to let your gene-data to be shared for scientific research purposes. Given their huge database, 23andMe works together with research organisations, such as universities to provide these non-industry organisations with valuable gene data that wasn’t as structured and available in this size before. Of course, all this individual data will be made anonymous.
With regards to your concern on reliability, she argued in the same interview that given the size and scale of the 23andMe database ensures accurate outcome for everyone. But I do understand concern. I think it’s good that you do point out the difference in statistical truths and in-depth analysis.