When I learned that Dr. Hannah Fry was going to speak at Mendix World, I was eager to hear what she had to say.
Hannah Fry is an associate professor in the Mathematics of Cities at the Centre for Advanced Spatial Analysis at UCL. She studies patterns in human behavior - particularly in an urban setting. I was first introduced to Fry’s work through her TEDTalk, The mathematics of love, and I read her books soon afterward. She quickly became one of my role models, showing me that I could pursue a career in STEM (science, technology, engineering, mathematics), too - a career path I’m now fully exploring and enjoying at Mansystems/FlowFabric as fullstack developer for SMART Digital Factory Tooling.
With Mendix World going virtual this year, Dr. Fry’s public keynote was changed to an exclusive workshop with 30 limited seats. Four Mansystems/FlowFabric members were selected to attend this special event for their contributions to the Mendix community. When my colleague, Andrej Gajduk, who was Maker of the Month in 2020, learned about my passion for Fry’s work, he offered me his spot, which I gladly accepted.
Dr. Fry introduced herself as a mathematician who uses data to study the patterns of human behavior. In her view, the world exists in two realms: there is the physical realm, where we say and do things, and we have a layer underneath, the world of mathematics. Fry believes that this second world, which you can’t see at the surface, shows much about how the world works.
Fry kicked off with a slide full of different keywords: serial killers, race cars, COVID-19, cows, and nuns, referring to stories we had yet to hear. At first glance, all these topics didn’t have anything in common, but as the workshop continued, we learned that everything was about data and that data isn’t always what it seems. Fry engaged our moral and scientific compasses, starting with the following thesis, pulled from a story about cows:
People who own smart speakers should inform guests before they enter their home to let them know they may be recorded.
With a master’s degree in history, I’m often divided about cases of devices and services (and the companies behind it) that collect data. On the one hand - and this is the argument that came up most often in the discussion around the thesis - the possibility that data that we give away so voluntarily could be used for something bad and might backfire. So, on that note, you should let your guests know before they enter your house, so they can decide for themselves whether or not they want to give away their valuable asset (data) for free. On the other hand, the historian in me loves data. Data can give scientists wonderful insights into how people behave through time. Can you imagine how the study of history will evolve when so much is being recorded?! Which is why I want people to be themselves, and when you give them the knowledge they might be recorded, their behavior might change, corrupting the data.
Fry also briefly highlighted that the way most voice assistant devices are programmed to be female as the default female runs the risk of strengthening sexist stereotypes.
The second story Fry shared was the Nun study. In that study, 678 nuns participated in research investigating the causes of Alzheimer’s disease, trying to match the nun’s cognitive ability in relation to their age. The nuns also voluntarily donated their brains after their death so that the researchers could determine whether signs of dementia were present in the physical part of the brain. They weren’t. However, due to a seemingly unrelated data source and an algorithm that could connect the two different sources, the researchers found that the chance of developing signs of dementia could be traced back to children’s use of language. (1)
Predicting the future with the help of data also played a part in the story about serial killers. By using algorithms that can compare different data sets, new patterns can emerge. In this case, data could help to pinpoint a serial killer’s identity or location. (2) So there is great potential for algorithms that to improve the probability of showing, solving, or even detecting crime before it happens. However, Fry pointed out, there’s a big ‘but’: the code is only as good as the programmer who writes the algorithm, meaning that bias is also programmed into scripts, which can have disastrous consequences for innocent people. Especially when you realize that algorithms are still far from predicting the future 100% correctly.
The last story was about racecars. Fry asked us to imagine that we were a racing team. The car’s engine had already failed a couple of times before, and there was a suspicion it had something to do with cold temperature. It was going to be 40 degrees Fahrenheit on the day of the race. We had to decide if we should race or not. Oh, and if we didn’t choose to race, a lot of people would be angry because they’d invested tons of time and energy in the project. Fry showed us a graph charting the number of engine failures and the temperature in Fahrenheit. The failures happened anywhere between 54-75 degrees, so it seemed a bit random. Like almost everyone else who participated in this study, our group chose to race. However, we based our decision on incomplete data. No one of us asked if we could see the data on races with successful outcomes. As Fry showed us next, when we added this data to the graph, it was clear beyond a shadow of a doubt that no race had ever been won below a temperature of 65 degrees. To race or not to race? The data used in this Carter Racing Case was based on a real fatal incident, the Space Shuttle Challenger disaster. The Challenger team decided to launch the shuttle, causing the shuttle to break apart 73 seconds into its flight, killing all seven crew members. (3)
In our interactive workshop with Dr. Hannah Fry, we looked at seemingly different stories that all have one truth in common: data is only as useful as the person who interprets or applies it. Whether it’s the first story about turning a blind eye to potential misuse of data we give away, but which can be useful if handled ethically and correctly, the nun study on connecting data that, at first glance, is unrelated, but proves to have unexpected correlations, the serial killer story on being critical about programmed bias and the successful prediction rate of asking algorithms to help you forecast the future, and lastly, the racecar example, highlighting how we should always keep in mind that the data we’re presented might not be all the data, and we shouldn’t be afraid to ask for additional information when important matters (or even lives!) are at stake, the message is the same: humans should be present.
“But,” said Fry as she ended the workshop, “even though you might have all the data in the world, sometimes we need better data that does not exist yet. Because we can’t know everything.” And as she says in her book, Hello World, one thing is certain, “in the age of algorithms, humans have never been more important”. (4)
(1) Hannah Fry, Hello World: being human in the age of algorithms (2018) 90-92.
(2) Fry, Hello World, 141-146.
(3) If you want to learn more about the Carter Racing Study and the Space Shuttle Challenger disaster, David Epstein included this case in his book Range: how generalists triumph in a specialized world (2019) 243-250; or read the in-depth study by Diane Vaughan: The Challenger Launch Decision - Risky Technology, Culture, and Deviance at NASA (1996).
(4) Fry, Hello World, 202.