10 things I learned about how to have a positive impact as a data scientist

getal-10-kunstwerk.jpg

In the first two parts of this mini-series, I outlined how you can develop your own approach for having a positive social impact. A key element is doing experiments to find out what works for your specific values, skills and situation.

Here are 10 things I learned in the last 15 years of conducting such experiments.

  1. Leaving the business world for the nonprofit world might be one option. As someone who has worked in and with nonprofits in different technical roles, I would not recommend it yet on a large scale. There aren’t many positions available for data scientists yet. And those that exist may not yet provide much opportunity for meaningful impact.

  2. Being picky about who your work for is another option. There may be companies who have ambitions that align better with your specific idea of „meaningful impact“. Currently, I would say I have found such an employer. I wouldn’t say there are too many of them, unfortunately. But your values and expectations might differ so it is worth looking.

  3. Improving the ethical practices in your industry, domain and company. Concerns like privacy abuses or biased algorithmic decision making need competent and impact-minded people inside the business sector. People who advocate for and implement ethical standards. This will definitely be an uphill battle, but if you can pull it off, you can enjoy a large positive impact.

  4. Volunteering. There are different ways to approach this. One is volunteering with pro bono data science organizations such as Data Science for Social Good (DSSG) Berlin, Correlaid, and Datakind. By volunteering long-term in an individual nonprofit. Or by volunteering for different organizations on a per-project basis. Avoid purely technical projects that don’t work directly with nonprofits or other civil society organizations (e.g. setting up your own „green field“ open data project without working with future users).

  5. Having significant impact is difficult as a data science volunteer, however. Short-term engagements such as participating in hackathons or individual projects are fun. But not sustainable. Long-term engagement with nonprofit organizations is necessary to prepare projects and follow up until a solution is actually useful.

  6. Many nonprofits are not ready to start data science projects and need help with identifying and scoping potential projects. Finding nonprofits that are able to benefit from data science projects is one of our main challenges in DSSG Berlin. (Finding data scientists eager to volunteer, however, is super easy!)

  7. Choosing to volunteer your skills for an individual organization is another option. This allows you to drive projects from conception all the way to the point where your solution is used. But finding an organization ready to host data science volunteers requires a very good network. You are not likely to find data science roles listed on volunteering portals or other places accessible via Google.

  8. If you choose volunteering as a path, regardless of your approach, you will need to learn how to be an effective volunteer. Start by learning as much as you can about the challenges nonprofits face. From fundraising to reporting on their impact, from minimizing financial and legal risks to staffing. Learn how they deliver on their programs and strategy. Data science can help with many of these, but you need to learn how these are done and the language that practitioners in these areas speak.

  9. Then, learn how to work with a very limited tech stack. Nonprofits usually won’t afford adding new tools to their stack. Instead, you will have to learn how to integrate with whatever stack the organization is currently using.

  10. Finally it helps to pick organizations that understand their impact and are well-managed. Ask for their „Theory of change“ or logic framework, for example. If they are a small organization starting out, you can help them work towards these goals instead.

Those have been my main insights from my own experiences over the last 15 years and from hearing other data scientists’ stories. I would love to hear more perspectives: What has been your experience? Please let me know via Twitter or email!