Excella’s own Aaron Pujanandez recently engaged in a thought-provoking conversation with John Gilroy on the Federal Tech Podcast. This podcast, where John regularly convenes with prominent figures in the industry, serves as a platform for deep-diving into the intricacies of federal IT. Throughout this discussion, Aaron and John explored an array of topics, including the basics of data science, applications of data science in the private and public sectors, and AI’s impact on the data science field. Here are some highlights from the conversation:
Ethics in Data Science
Ethical considerations are paramount in data science, particularly when decisions derived from data may have life-altering consequences. Aaron and John delved into why it’s critical to have a human in the loop, and the need to maintain ethical considerations when developing models.
“Particularly in these life and death situations, and I would say, particularly in most public sector situations, where even if they’re not life and death, there is an individual on the other end that is going to receive some outcome that may be negative. I think the human in the loop is key. Ethical considerations [that go] into model development are also things that you need to start thinking about.”
How the Federal Government can Derive Value from Data
Aaron brought up some key reasons why the federal government has a had time getting value from their data. One of the reasons is the sheer amount of new tools that enter the market at an ever increasing rate. Government agencies want to keep up with citizen needs and demands and they need new tools to do so. However, it’s difficult to carve out time to set up the right foundations. Government leaders should be asking questions like:
“What does the data need to look like? How is the infrastructure going to support this? Have I actually gone and done enough research on these specific decisions and the inputs to them that I need to be able to derive value from these things?”
Balancing Security and Data Access
Achieving the delicate balance between data security and accessibility is a critical concern. Aaron expressed support for security and access restrictions on sensitive data, emphasizing the importance of only using the data necessary for the mission. When asked how federal agencies can utilize the massive amounts of private data, Aaron replied:
“I happen to be a fan of most of the security and access restrictions to a lot of the data, not all of the data, but a lot of the data that exists within the federal government simply because a good number of it is sensitive. It has a lot of details that really should not even be shared among individuals. There’s a number of different ways to systemically to allow somebody to get the value from data insights without actually having access to the underlying data itself.”
Communicating Complex Data to Non-Technical Stakeholders
In the ever-evolving landscape of data analytics, effectively conveying complex information to non-technical stakeholders is an art form. When asked how Aaron conveys extremely high-level of data analytics to those outside his field of expertise, he replied:
“We use a lot of human centered design techniques when we’re actually building up those decisional support tools… [We] really [focus] on what people are looking at and how they’re making the specific decisions that they’re driving for.”
To build off of that idea, Aaron described a process he called “Dashboard for Your Dreams:”
“We get a bunch of government contract workers in a room. And we literally just start driving around, ‘if you could have any dashboard that would help you do your job, what would it look like? What would be the types of information you would have there?’ It’s always a give and take, because when you start that way, and then only build that, you’re sometimes missing the value of data. So it has to be a continually evolving conversation.”
Looking Forward
To wrap up the interview, Gilroy asked what tool Aaron would like to see developed in the data analytics field, to which he said:
“There’s a lot of divergence in how you actually develop AI software as of right now. So you can look at things like JIRA support tickets. You can look at Model Management regimes, but there’s nothing really that ties them together that does a good job of being able to provide that decisional support kind of one view thing. You have Tableau, where you can go in and see a dashboard, but there’s really no way to see the overall health of the data science industry or the data science practices within your specific company.”