Algorithms and Bias
Algorithms and Bias
image source: tctmd.com
This week, I decided to learn more about algorithms, how they work, and how they affect what people view online. Before digging into some resources, I knew that algorithms could be biased but didn't realize to what extent. The phrase "it's just math" came up a lot in the resources I read and listened to, when in reality algorithms are not "just math." Algorithms are created by human beings, which means that technology, specifically algorithms, is not untouched by human bias.The resources I interacted with this week explained bias in algorithms, but also revealed some of the benefits of using algorithms.
Algorithms are an important topic to understand because as educators, we need to teach our students to access reliable, accurate information. We also need students to understand the concept of bias and how to fact-check and question the information they consume. This relates to my current learning space because I often have students use the Internet to complete research. Although we do discuss credibility in sources, I have never talked to students about bias in algorithms. Some of my students will get information from the first website that appears after a search instead of browsing several resources and then deciding which to use in an essay or project.
One of the most surprising discoveries from this week was how much faith people put into algorithms in regards to accuracy and unbiased information. According to "Why Search Engines Aren't Unbiased With Safiya Noble", many people believe that "...the mere presence of complex math assures accuracy and that technology has an authority we should respect, if not obey" (Conover). Adam Conover goes on to give examples of algorithms giving wrong information to users. One such story is about a marathon runner who became upset that the course wasn't the correct length when in fact the FitBit they were wearing measured the wrong distance. The FitBit is not always accurate when counting the number of steps or the exact distance of a run, but people are ready to believe that it is always accurate. Conover goes on to describe people putting their trust in technologies such as GPS and self-driving cars without questioning how well they work or how they may be flawed. Humans created these technologies and humans make errors and have biases. When it comes to the topic of search engines, humans created the algorithm that can lead to bias in search results. Some companies even pay search engines like Google to get their content to show up first. Therefore, algorithms are no longer just "math", but influenced by humans and their interests.
Steven Bellovin's article "Yes, 'algorithms' can be biased. Here's why.", also discusses this idea. Bellovin mentions instances when algorithms became discriminatory against women or used racist terms to describe people. As also mentioned in the podcast, I learned about the idea of GIGO (Garbage In, Garbage Out). Essentially, this idea is that the quality of the output depends on the quality of input. If people input biased or inaccurate information, that is what the system will output. I've never heard of this phrase before, but it is a great way to describe the phenomenon of "fake news" and unreliable sources.
image source: researchgate.net
Although algorithms appear to have many negatives, there are also some positives. In the article "Code-Dependent: Pros and Cons of the Algorithm Age", authors Lee Raine and Janna Anderson mention that "...everything people see and do on the web is a product of algorithms." They also go on to discuss how algorithms are extremely helpful in automating processes, such as Netflix recommending a movie to watch or the Amazon app recommending products to a user. At this point, algorithms help people to manage their everyday tasks and are intertwined in the decisions people make. Algorithms also "...make sense of massive amounts of data" (Anderson, Raine). After listing some positives of algorithms, these authors also return to the idea that algorithms can cause inequalities and will reflect people's biases. Therefore, the article suggests the need for algorithm literacy, transparency, and oversight.
After reviewing all of this information, I am compelled to begin teaching algorithm literacy in my classes while also teaching about finding credible research sources. Even though I do find algorithms sometimes useful, such as finding accounts on Instagram I'd enjoy following or having products I'd buy advertised to me, I now understand how algorithms can also put people into a box.
Resources
Bellovin, S. (2019, January 24). Yes, “algorithms” can be biased. here’s why. Ars Technica. https://arstechnica.com/tech-policy/2019/01/yes-algorithms-can-be-biased-heres-why/
Conover, A. (2020, January 21). Why search engines aren’t unbiased with Safiya Noble. Earwolf. https://www.earwolf.com/episode/why-search-engines-arent-unbiased-with-safiya-noble/
Rainie, L., & Anderson, J. (2017, February 8). Code-dependent: Pros and cons of the algorithm age. Pew Research Center: Internet, Science & Tech. https://www.pewresearch.org
Hello: I found your post very information but particularly like the following sentence: "After reviewing all of this information, I am compelled to begin teaching algorithm literacy in my classes while also teaching about finding credible research sources."
ReplyDeleteEducation is indeed the best method to shelter people from aspects of algorithm use which can be detrimental to users. In the following video which I found during my research, I learned that for "7 billions users, only about 70,000" actually know the workings of search engines and algorithms. If we are going to use social media and emerging technologies, it is our duty to educate ourselves about the pitfalls and advantages that they offer.
Thank you!
P.S: Here the link to the video:https://www.youtube.com/feed/you
I appreciated that you had a balanced view, Kate. Everything has a positive and negative side to it. When we teach students about algorithms its not to get them *not* to use them- as you stated, there are some positives. But instead, it's to help them understand that they have choices.
ReplyDeleteI think you really did a good job of explaining what algorithms and how they are tricky. That diagram of the input confirming biases was a great choice to share! I think teaching how algorithms work is an important aspect of information literacy and basic technology. I just want to echo some of the other comments that algorithms are not all bad. They are a tool, but being ignorant of their affects and how they work is when they can have potentially insidious outcomes.
ReplyDeleteI found your Fitbit analogy particularly interesting. I think it completely demonstrates how much weight we put on the information being presented to us being accurate. While Fitbits might encourage more activity which is not a bad thing, other false information could lead to negative consequences. I also liked the diagram you included because it clearly shows why views can be reinforced by social media. Personal data makes it clear to social media what beliefs a person has, and then those algorithms feed them more information that reflects their personal views. I agree with you that this is something that should absolutely be taught in classrooms. Do you see digital literacy turning into standards that all schools are required to teach in the future?
ReplyDeleteWhen you brought up the point of the GPS not being accurate, it reminded me of a “news” series (Note: A bit not safe for work due to swearing. The Host also has a Tech Q & A on an excellent side channel) I watch every week where they discuss people’s weird crimes and shenanigans that get printed in the news. At least twice, individuals have followed their GPS and drove their cars into bodies of water. Boston & Hawaii. It just throws me that these individuals were getting close to a body of water and still believed that the GPS was giving them the correct directions.
ReplyDeleteThe Garbage In and Garbage Out remind me of the time an AI Vtuber (Virtual Streamer) was banned from Twitch for denying the Holocaust happened. I am more annoyed as the developer didn’t think this would happen when they allowed the AI to roam the internet.
My problem with algorithms is that spam tagging exists, and there are little to no repercussions for doing it. Individuals add popular tags to their random posts, filling the tag search with garbage. If I remember I got a warning on an account I had; I don’t remember which site it was for mass flagging too many posts in a row. I was like, this person is spamming the tag, and I’m telling you it’s incorrect to be here, but I’m getting in trouble for it?
YouTube “News” Site: https://www.youtube.com/@RadioDeadAir
Boston : https://www.boston.com/news/local-news/2023/09/25/shades-of-michael-scott-doordash-driver-charged-after-following-gps-into-water-in-middleton/
Hawaii: https://www.businessinsider.com/tourists-hawaii-gps-drove-car-into-water-2023-5
AI streamer : https://kotaku.com/neuro-sama-twitch-vtuber-ban-holocaust-minecraft-ai-1849977269