When did you last fool an algorithm?😉
When did you last fool an algorithm?😉
Hello!
Seven weeks ago, I started an experiment. I shot my own 4-minute whiteboard videos on leadership in data science. While I’ve enjoyed recording each episode, I learned a LOT more about framing videos, getting the lighting and audio quality right!
If you’d like some tips, let me know. You can watch all the 7 byte-sized episodes, here.
This newsletter will take you about 4 minutes to read.
I. Spotlight: When did you last fool an algorithm?😉
In the 1920s, a landmark experiment was conducted in Illinois to study how factory conditions impacted worker productivity. As researchers increased the levels of lighting, the productivity of workers improved.
When researchers thought that they had their answers, they noticed something unusual. Worker productivity started improving with subtle changes in many other factors.
They realized that something else was driving all the gains.
The researchers then found out the real reason. It was the very fact that the workers were being observed!
When workers knew that they were being studied, they changed their behavior. Nothing else seemed to matter.
This came to be known as the Hawthorne effect, named after the Western Electric plant where it was first noticed. We can see this playing out in many areas of our lives today.
What would people do when algorithms watch them 24x7?
Today, algorithms are everywhere and they study us continuously. They use live data feeds to observe how we think and react.
Often, the decisions that these algorithms make have huge social and economic consequences.
Which friend should Facebook recommend? What post should Linkedin promote? Is a customer eligible for discounts? Should a suspect go to jail?
People change their behavior when they know they are being watched. And, the higher the stakes, the more pronounced is this change.
Don’t we already notice this shift in behavior?
Commenting on Linkedin posts to have them shared widely. Manipulating steps on the Apple Watch to fool friends, or to trick the insurance company. Abandoning online purchases in order to get discount coupons. The list goes on.
Some of these manipulations are subtle and benign. But others are just unethical. And, some even have a profiteering motive.
What does this mean for your company?
When organizations build predictive algorithms, they get excited about the initial results. But, they are ill-prepared to respond to systematic manipulations, or to the continuous shift in consumer behavior.
What should leaders do to equip their companies to respond?
There are several strategies to tackle this problem, but I will share just one for this newsletter.
Invest in deeper behavioral skills
When the quick wins from superficial consumer signals evaporate, you must dig deeper. You can detect the gaming of algorithms by looking for deeper behavioral patterns. Those are the subtle human responses that are hard to manipulate.
You might already have data scientists and domain specialists on your team. Adding behavioral experts and social scientists can give you this capability.
This is quite like getting a new pair of glasses to see through the foggy vision.
These experts can figure out why people behave the way they do. They can also find out when the signals are being tampered with. Keeping social scientists in-the-loop for decisions will improve the reliability of your outcomes. It will also make the decisions more humane.
Is your organization taking steps in this direction? How do you think this will play out in the future? Reply and let me know.
II. Industry Roundup:
1. Compilation of MIT Data Leadership Event talks
Last month the MIT CDOIQ brought together Chief Data Officers and Industry Thought Leaders. Here’s the library with over 60+ talks. Do check it out.
->Â Watch the videos
2. These weird photos show that AI is getting smarter
5 minutes | MIT Tech Review | Karen Hao
Ask a kid to draw ‘a cat on the wall’ and she wouldn’t have any trouble. But, this was super difficult for AI, until recently. This article shows how far algorithms have progressed and its implications for the future.
->Â Read the Article
3. Why Culture is the Greatest Barrier to Data Success
7 minutes | MIT Sloan Management Review | Randy Bean
The greatest barrier for organizations to become data-driven is business culture, not lagging technology. Check out the 6 ways to power your business transformation.
->Â Read the Article
III. From my Desk:Â
1. Whiteboard video: How do you identify the best Data Science Projects?
How do you build a laundry list of projects to pick from? Here’s a combination of business-driven and data-driven approaches that we’ve evolved at Gramener.
->Â Watch the video (4 min)
2. Article: 9 Ways your remote Leadership might be hurting your Culture
9 min | The Enterprisers Project | Ginny Hamilton
Can your leadership style backfire with a team that’s now fully remote? Watch out for these 9 bad habits compiled by technology leaders.
->Â Read the Article
3. Upcoming Talk: Saving lives by applying AI on Satellite Imagery
Combine the advances in computer vision with high-resolution satellite images and it opens up new possibilities. In this talk, I’ll show a live demo on how we used both to help save lives.
->Â Register for the free event
Missing that offsite meeting or those flight journeys?! Check out this unbelievable ‘flight-to-nowhere’ that was fully booked in Taiwan, thanks to the pandemic!😅
Yours,
Ganes.
PS: Did someone forward this to you? You can subscribe here.
My Website | LinkedIn | Twitter | YouTube