Dear Trish, I don’t really know exactly what an algorithm is. Also, are they good or bad? I’m so confused.
First and foremost, thank you so much for this important question (and for being honest and open to learning!). I want to begin by validating your confusion: while you likely interact with algorithms – whether simple or complex – each and every day, the concept is rarely explained to users in a way that’s simple and accessible. So don’t feel bad – lots of folks, especially youth, can’t exactly say what an algorithm is. And don’t fear: in this post, we’ll not only define an algorithm, we’ll discuss where they’re employed in our everyday lives. You also raise an important subjective question: what do algorithms mean for our society? While it’d be hard to label algorithms as “good” or “bad,” there are absolutely pros and cons to their applications in different parts of our world. With that said, we’ll also discuss some of the backlash algorithms have seen in the last few years (and you can be the judge!).
Let’s get into it – and begin by understanding what, exactly, an algorithm is. Everyone has their own definition, but personally, I love David Malan (a Harvard professor)’s simple take: “in computer science, an algorithm is a set of instructions for solving some problem, step-by-step.” As Professor Malan explains, we humans have algorithms too, whether it’s for counting people in a room or even making dinner. When we generally refer to “algorithms, ” though, we’re usually thinking about instructions executed by a computer. What are these computer instructions meant to do? It’s put best in this video: an algorithm “enables a computer program to put together different sources of information and generate a result.” As the video goes on to explain, we have a lot of data in our world, so when we ask computers for the answer to some question (e.g. make a Google search), the computer program needs a set of instructions to figure out what information is relevant and what the right answer is. And that’s where algorithms come in. All that’s to say: when you hear the word “algorithm,” think “instructions for info. collection and problem-solving.” It’s that simple.
Okay; we have our definition. Where do we see algorithms in our everyday lives? In short: everywhere. To be more specific: search engines, dating apps, social media platforms, and on, and on, and on. When you search something up on Google, for instance, there are algorithms – sets of instructions – that, based on listed sets of criteria, generate what results are returned to you, and in what order. Same thing with social media platforms. As you’ve probably noticed, when you’re on Instagram, what pops up on your home feed isn’t necessarily your friends’ most recent posts. Based on Instagram’s algorithms – its instructions – it could be a specific account’s posts, or posts that feature a specific type of content (for instance video), or even an ad. Algorithms aren’t just confined to the content, we see, though; increasingly, algorithms are used in everything from hiring to assessing where law enforcement should be patrolling in your neighborhood. Again, all that’s to say, you probably interact with algorithms on a daily basis (potentially without even knowing it!).
Okay – now, let’s get to the “juicy” stuff. What, if anything, has gone wrong with algorithms, and what have folks been trying to do about it? Well, in recent years, there’s been a lot of backlash and concern over biased algorithms, or algorithmic bias. What’s algorithmic bias? Algorithmic bias refers to algorithms that are systematically unfair to specific groups of people. How would that happen?, you might be wondering. Well (without getting too technical), a lot of sophisticated algorithms actually “learn” their instructions from data in the real world. For instance, a new hiring algorithm, designed to replace interviewers, may “learn” their instructions about who to hire and who to reject from data about which candidates were successful in the past at a company. But what if that company historically has had issues hiring and retaining women and people of color? Well, that bias might get passed on to the algorithm – because the data seems to say white men were “more successful” at the company, the algorithm may only offer jobs to white male applicants. If you’re thinking “uh oh,” right on.
Algorithmic bias has gotten a lot of attention as we’ve started to use algorithms in many parts of our lives, including areas where decisions made by algorithms can be life-changing. Hiring is a great example, as are recidivism risk scores. Recidivism refers to when a criminal reoffends, or commits another crime. Across the country, lots of courts rely on computer algorithms to produce a “risk score” that a criminal might re-offend, and judges often use the score to determine whether a criminal can get out on bail, or how high bail is set at. But over the last few years, researchers and organizations have argued that the instructions the algorithms have learned are biased, making it more likely to falsely flag black defendants. That, of course, is a huge problem – racial inequities in our society are being exacerbated by algorithms.
You might be wondering – have folks tried to “fix” algorithmic bias? Well, a key way to address algorithmic bias is to make algorithms transparent: that is, to understand how they “learn” their instructions and what those instructions are. (In the tech world, many people refer to this as “getting inside the black box.”) But there are limitations. Some companies don’t want to make their algorithms transparent, and sometimes, depending on the technology being used to build the algorithms, it can actually be really hard to understand how an algorithm has come to a conclusion (even if you want to). (I know – sigh.) It’s also not clear how to hold an algorithm accountable: if it doesn’t hire someone it should have, for instance, we can’t really “punish” it, can we?
I hope you found this post helpful in answering your questions about algorithms. Whether you now have other algorithm-related questions and want to keep the conversation going or you’re ready to pivot to another issue on your mind, don’t hesitate to share any Internet-related questions, thoughts, or perspectives here. Your question just might be featured in an upcoming TikTok/blog post! Remember, not only will you get some helpful advice, you’ll also help our Ask Trish community: other young people, just like you, wondering the same things. In fact, by sharing your thoughts, you might just inspire them to ask a question too! So take 1 minute, and fill out the form. I can’t wait to hear from you!
Until next Tuesday,
Trish