(September 7, 2017) Today was the first session of our teen course “Our Algorithmic Culture.” My goals for today were to stimulate people’s imagination and curiosity, give students a sense of what an algorithm is, and provide some baseline cases and examples to keep returning to. I’d like to fill you in on the questions I asked that guided our discussion today.
>>>What is an algorithm? The students defined it as an input and output with a certain way of solving something. They brought up and asked some questions about the Pandora algorithm (specifically it’s “randomness”), which we will return to during another session.
>>>What are some common algorithms and how do they work? We worked through conversion from Fahrenheit to Celsius and also prime factorization. Then I showed the students an article in today’s Philadelphia Inquirer about new Youth Poet Laureate Husnaa Hassim.
>>>If you wanted to create a website or app that provides its users with a poem every day, how would you be sure to provide poems that your users would like? The students said they’d ask users to click if they like the poem, watch for trends (feedback!), and survey users ahead of time for 6 things: what’s going on in your life, how was your day, genres/poets they already like/dislike, your religion, your core values, and your style of humor. A discussion emerged about what the right number of questions would be. “What does this have to do with algorithms,” asked the class. You will see, you will see!
>>>What would you do if you were in charge of a large, urban, struggling school district and you needed to raise student performance? The students brainstormed and debated. I asked “who is easy to blame?” for student performance. I asked what variables they would consider if they had to rank teachers to weed out ineffective ones. The students said they’d ask kids, sit in and observe, evaluate teaching performance, and perhaps give teachers a sample lesson to teach. They grudgingly said that if money were an issue, they may have to use student testing data, but that they wouldn’t want to. I then shared the story of teacher Sarah Wysocki’s unfortunate experience with the teacher performance algorithm “Impact” described in detail in Cathy O’Neill’s Weapons of Math Destruction (WMD).
>>>Do you care how the temperature algorithm works? Only one student raised her hand.
>>>Do you care how the prime factorization algorithm works? That same student raised her hand, and a few put their hands up a few inches, but most kept their hands down very low.
>>>Do you think people care how the Impact algorithm works? “They should!” declared D, with many nods of agreement. (I’m hoping that by the end of this course, people will not automatically accept mathematical algorithms – whether pure math or applied – on faith.)
>>>What is your algorithm for packing your lunch?* The students brainstormed a list (which would, of course, vary by student) of the variables they would consider: food availability, hunger, what they’re doing that day, nutrition, taste, portability, quantity, whether sharing, dietary restrictions, and variety. (Just an aside here: I found it interesting that no one mentioned cost. I wonder how this list would differ had parents written it.)
>>>Could you write down your algorithm so that someone else could pack your lunch for you? Students weren’t sure. J began to write her algorithm. Others began to discuss potential difficulties: things that can vary day by day, the skill of the person preparing it, mood, birthdays, availability, seasonality, tiredness. We then discussed the difference between a formal and informal algorithm and also what it means for an algorithm to be trustworthy. “So you’re saying that the algorithm has to be dynamic; it has to react to changing conditions?” Most said yes. J, however, posited that her algorithm didn’t have to be dynamic, that it would work every time. M seemed to agree, and began proposing his own algorithm. J read her algorithm to the group. M and J posited that it might be possible to forsee every variable. People started poking holes in their algorithms. “You’re behaving exactly like mathematicians,” I remarked. “One person posits a proof, and other mathematicians try to find its flaws.” The discussion continued. The debate focused on whether an algorithm must be dynamic in order to be trustworthy. This was a debate I definitely didn’t anticipate when I was planning this session, so it was thrilling to watch.
We were then out of time, to be continued next week.
*The lunch packing algorithm example is based upon O’Neill’s dinner prep example in WMD.