Data Science for Public Policy: How I Am Plowing My Way Through Imposter Syndrome
Three years ago, if you told me that one day I would use python to analyze AI policy and make Guido van Rossum chuckle, I would think you are crazy.
Three years later at PyCon 2019 in Cleveland, that’s exactly what happened.
I was by no means a tech person. I was trained as an economist (read: stats nerd), but somehow for the past three years I’ve been writing analysis on deep-tech fields including AI and 5G.
what this means is I have very severe imposter syndrome.
What I hope to achieve with this post is not #humblebrag (ok, maybe a little happy dance) but to share with you all the struggles I had and am still experiencing on a daily basis and to reassure a fellow researcher somewhere feeling that he/she is faking it all the time, you are not alone.
First of all, just look around this field — everyone in the room seems to be a former policy maker, DoD expert, ex-diplomat, or at least professor in Computer Science with tenure under the belt. Also, look at the subject we are debating here: artificial intelligence, 5G, national security, tech competitiveness — the stakes seem too high to mess up.
hey man, you sure you aren’t calling the wrong number?
In the industry of ideas, I struggled to find my voice because I did not have years of experience in the department of defense or in national security, nor can I tell anecdotes about being on the negotiation table when China joined the WTO.
But taking a second look around the room, I realized like many younger folks in this area 1) although we don’t have singular data point from personal experiences, there are tons, tons of data out there that’s readily available. Being a former journalist myself, I recognize the inherent danger of outlier data points especially in the area of security and the value of a more holistic approach; 2) these policy subjects — AI, 5G, competitiveness — are incredibly new. Count your fingers (and maybe your toes); that’s about the total amount of researchers in any of these sub-fields. Instead of thinking ‘omg what if I am saying something that’s completely wrong and laughable’, try ‘you know what, whatever I have to say, as long as it is based on honest research, I’m pushing the boundary of the collective understanding on this issue.’
Thus I turned to data science two years ago shifting to python from SPSS and Stata, hoping to leverage its interpretability to a larger audience. I also started presenting code and research at tech gatherings hoping to get feedback, critics, and more importantly help with technical difficulties because trust me, stack overflow can only carry you so far.
Then I got really lucky. At one of the presentations, some judges from PyCon 2019 were there. We talked briefly after my presentation and a month later he emailed me, offering a spot at PyCon Startup Row as one of the sixteen teams nationally. And thanks to the awesome Project Alloy, whose mission is to build a more inclusive technical community, I made it to Cleveland in May (shoutout to Brooke and K!)
I talked to a lot, a lot of folks at PyCon confiding in them about my anxious situation. And guess what?
everyone has imposter syndrome!
We are talking about engineers at google who is in charge of network stability, data scientists at a leading bank who built the credit risk model, engineers who built the pricing algorithm for a ride-hailing company. Besides the reassurance, I also learned a couple tips to deal with my anxious self:
- do honest work
There is a lot of ways to slice and dice data, any data. When in doubt, do a significance testing or simply present the alternative for discussion.
2. present your work honestly
Everyone is constrained by time, capacity or both. While it is tempting to make a statement that has a higher chance to be quoted, I will not go there if my research doesn’t stretch that far. This not only helps with alleviating imposter syndrome but also contributes to a rational debate on hot-button issues.
3. if you can, share your work
I try to share as much code and streamlined data as I can — for both selfish and not-so-selfish reasons. It is a blue sea of data in tech policy out there. I want to signal and share the part I’ve covered so that as a community we can tackle more ground. Also, if I reached an incorrect conclusion exposing my work to the public makes it easier for people to point it out, allowing me to learn from my mistakes (be gentle please...)
These tips have been working well so far (the next couple sentences are the happy dance part. Feel free to skip) — My research got picked up by DC-people’s fav newsletter Axios, I gave an interview to the Bulletin of the Atomic Scientists, and a Special Report of The Economist cited my article on AI (Google me, Justin! the econometrics professor who gave me A-!)
There is a looooong way ahead. Averaging twice weekly I’m on the phone with a journalist or an expert and want to find a hole to hide in.
But some folks have touched upon the imposter syndrome topic here and there so I wanted to share my thoughts and say:
yes, there is a way through it.
Find me on the internet and poke me, if there in any way I can be of your help to plow through imposter syndrome or if you are a fellow nerd in public policy & data science or just wanna say hi.