Creativity in Data Science: A Southwest Analyst's Dog Button Board
In a session at the Data Science Innovators Conference 2023, Southwest Airlines analyst Mary Stimson argued that data science can be an art form and showed how curiosity, learning outside your domain and failing fast shaped a project that tracked dogs using communication buttons. She also described her team's 80/20 rule for innovation time.
By The AI Enterprise Magazine Team4 min read
Key takeaways
- An artist is behind the model. Stimson said data science can produce art because people build the models.
- Do weird things. Projects outside work help build skills and curiosity.
- Fail fast. Her first dog button board worked and misfired at once, and she iterated.
- Provide information, not answers. People must interpret and sometimes override model output.
- Make room for innovation. Her team plans at 80 percent capacity to leave 20 percent for margin and projects.
In a session at the Data Science Innovators Conference 2023, Mary Stimson, a big data analyst and data application developer at Southwest Airlines, argued that data science can be an art form and described a personal project that illustrates how creativity works in practice: a connected board of buttons that lets dogs "talk" while logging what they press. Stimson, whose background is in accounting, advertising and ministry, said the talk was about where she hopes the industry goes more than how she got there.
Can data science produce art?
Stimson opened with her brother, a musician, for whom she built a dashboard in Google Data Studio that combines data from streaming and music-recognition platforms in different countries. It shows him where his music succeeds, which tempos and genres perform best, and helps him decide where to tour.
She said labels use similar data to decide whom to sign, and sometimes not on taste. That raised the question of whether data-driven decisions are still culture. She noted that music is itself "an algorithm" of rules that are often broken, which makes it easier for AI tools to imitate, and that her brother worries about tools that generate music in his style from a Spotify link.
Her answer was yes, with a condition. If culture is human intellectual achievement and art is the application of human creative skill and imagination, then models can create art "because there's always an artist behind the curtain," she said, pointing to the data scientist or developer.
What does a creative data science project look like?
Stimson's advice to creative people and to leaders fostering creativity was to do weird things, including things unrelated to work. She framed her dog project with the three tests for a patent, which she said are also a useful screen for innovation: something useful, novel and not obvious. She noted she has a patent pending.
The need she found came from speech-language pathologists who had adapted communication button boards for dogs. Seeing mostly anecdotal evidence online, including claims about dogs describing their dreams, she wanted to test whether any of it was real. Her professional skepticism, she said, is one of the things she likes about data science.
What steps did she follow?
Stimson organized the project, and her advice, into a short sequence.
- Encourage curiosity. She began during COVID, partly on work hours, with her leaders' support. She said a team that is curious will find even a fuel-efficiency problem fun.
- Think outside your domain. Before touching analytics she had to learn programming, woodworking, soldering and the Raspberry Pi to make buttons light up. Her team has also done woodworking to build structures for computer vision work in airports.
- Succeed and fail fast. Her first board produced a live dashboard of button presses, but it also misfired and once repeated "help" in a robot voice in the middle of the night.
- Do the math. Data science, she said, "separates us from magic," though the details were not the topic of her talk.
- Think like a human, again. Data science should provide information, not answers, because a person must interpret and sometimes override the output.
The final board was about eight feet long with space for roughly 72 buttons. She showed a video in which her dog Honey Butter pressed "that's all" after being told to leave it, and said that on a sample size of one the project validated her hypothesis and led to more.
How does Southwest's team approach creative work?
In the Q&A, Stimson said her finance-focused team uses little AI because finance groups often need models to be explainable to investors, and that she builds interactive applications in React, Dash, Plotly and Flask so users can run their own what-if scenarios and save a record of what was predicted at the time.
On guidelines for creative projects, she said her team tries an 80/20 rule: plan work at 80 percent capacity so the rest is available for mental margin, training and innovation. Asked about the most creative Southwest project, she described exploring how to tell where overhead bins are full, using the types of bags boarding rather than assumptions such as purse counts. She added that Southwest has remote data science roles in its IT department.
Where can data leaders hear more sessions like this?
AI Connect continues to program sessions on practical data and AI work. Its free 2027 webinar schedule includes 24 live webinars in four series, listed on the webinars page, and COLLIDE 2027 takes place Wednesday, October 13, 2027, at the Sandy Springs Performing Arts Center in Atlanta. Registration is coming soon; details are at COLLIDE.
Frequently asked questions
Is data science an art form?
Stimson said yes, because models are made by people: there is always a data scientist or developer behind the curtain.
How can leaders foster creativity on data teams?
She advised encouraging curiosity, letting people pursue projects unrelated to work, and using an 80/20 rule that leaves 20 percent of time for margin, training and innovation.
What is a good test for an innovative data science project?
She suggested the three patent criteria: useful, novel and not obvious.
Why does her finance team use little AI?
She said finance groups often avoid AI because it is not explainable to investors.
This article is based on remarks by Mary Stimson, Big Data Analyst and Data Application Developer, Southwest Airlines at Data Science Innovators Conference 2023 on April 12, 2023. More coverage: More AI Connect Events.
Filed under: People, Culture & Leadership, AI Use Cases & Applications
