The Tradecraft of Data Science

…for human intelligence

The term tradecraft is taken to mean high-level professional mastery of some discipline. Traditionally, that discipline is that of the Intelligence professional—the spies, the analysts, and the other personnel who support the collection of human intelligence. Informally, it has been called "acquired operational judgment."

So what, then, is the tradecraft of data science? Tools and tactics, yes, but also a cultivated intuition around how to collect, combine, and exploit data and technology systems to navigate physical and virtual worlds. Philosophically speaking, the tradecraft of data professionals is a style of mind to levy upon problem sets. Tactically speaking, it is the unique combination of exploratory methods employed to ask and answer questions of datasets.  

The question on which I have spent most of the previous calendar year is “how does one teach the tradecraft of data science?” How does one cultivate this tradecraft in the minds of a fundamentally new kind of practitioner—one who is able to design technically sound intelligent systems, ask discerning questions of those systems, and assimilate the responses (because in the age of AI, we the practitioners are still the arbiters of value).

This article explores the philosophical framing of the tradecraft of data science. It is also home to a few choice quotes.

Subsequent articles in this series may explore more tactical examples around data exploration, manipulation, and management. 

*

Beneath the mysticism and technical polish, the data scientist’s tradecraft is a choreography of skills: data engineering, statistical inference, machine learning, visualization, experimental design, and—perhaps most critically—interpretation. These practitioners understand that every dataset carries the fingerprints of human decision, bias, and design. The ethos of their craft is humility in navigating immense technical complexity—an acceptance that no model is ever the world itself, only a carefully constructed mirror of its patterns. 

This year I found myself on a journey to derive the requisite tradecraft equally from the skillsets of The Scientist, The Analyst, and The Technologist—the three archetypes in the narrative of modern technical inquiry and design. So what attributes and techniques, specifically, do each of these archetypes contribute to the tradecraft of data science? And how do we teach them in the age of generative AI, where “education will need to become about the training of perception and not the transfer of facts.” (Luke Burgis) 

*

The Scientist.

“Science is the only self-correcting human institution, but it also is a process that progresses only by showing itself to be wrong.” – Allan Sandage 

The Scientist’s contribution to tradecraft is commitment to the rigorous application of method—to forming hypotheses, testing with precision, and approaching datasets with both patience and skepticism. The goal is not to prove, but to understand. To teach this archetype is to train the mind to slow down—to value uncertainty as a signpost and to design inquiries and experiments that fail productively. 

At the center of classical empiricism and the modern scientific method are the tenets of structured inquiry, rigorous investigation, and the well-designed question. This is relevant even in how we ask questions of large language models (LLMs) and how we interact with our newly-developed agentic counterparts and AI delegates. Teaching scientific rigor and structured inquiry is about critical thinking, novelty, depth, and ensuring that the approaches to exploring observed phenomena are structured, exhaustive, and valid. From this archetype, we borrow on the importance of validity, explainability, and extensibility. 

*

The Analyst.

“...if fish were scientists, the last thing they would discover would be water… the context in which they are immersed.” - Joseph Tainter

The Analyst’s contribution to tradecraft is that of the interpreter—the translation of complexity into clarity in a way that facilitates decisionmaking. This has long been the task of intelligence analysts in the national security community. Where the scientist seeks truth, the analyst seeks meaning, distilling judgment and decisionmaking from the vapors of nuance. Teaching the analyst means teaching the art of narrative reasoning—how to construct arguments from evidence, and how to hold multiple, even conflicting truths in view until a coherent story emerges. How to weigh evidence. How to manage cognitive biases. How to evaluate fact while still taking into account the tiny fictions that the human mind imposes upon reality and human experience. 

In the skillset of the analyst, data is information. This information is munged and combined in unique ways to derive insight. The context is alive in the mind of the analyst. Recommendations are made and complexity is communicated using some mutually understood probabilistic language.  

*

The Technologist.

“Once you see the boundaries of your environment, they are no longer the boundaries of your environment.” - Marshall McLuhan 

A good friend and a very successful technologist frequently reminds me that in the context of intelligence analysis, there are only two ways to wield information against an adversary: 1) gain access to data that the adversary does not know you have, or 2) compute on the data in ways the adversary cannot fathom. Both are the work of the technologist.

The Technologist is fundamentally a builder—the architect of possibility. The technologist’s contribution to the tradecraft of data science is the engineering of systems that extend human cognition—the tools, frameworks, and architectures. This archetype thrives on experimentation and iteration, instinctively understanding that innovation begins not with certainty, but with a prototype. To teach the technologist is to nurture creative pragmatism: to make the abstract tangible, to approach seemingly intractable problems, to embrace constraints as design inputs, and to wield technology as a medium for amplifying human potential. The technologist reminds us that the future is not discovered—it is built.

*

Taken together, the skills of all three archetypes comprise the tradecraft of data science. It ultimately returns us to an ancient pursuit: understanding. We build systems to model and to predict, but the advantage lies in how we think about what those systems reveal. In this way, the data scientist becomes a new kind of intelligence professional—part scientist, part analyst, part technologist. In an age where generative systems can produce answers faster than we can frame questions, the rarest skill is not computation—it is discernment. Teaching this tradecraft is not about transmitting knowledge; it is about cultivating judgment. This is the work we must undertake as intelligence practitioners: thinking with rigor, building with intent, and navigating unstructured worlds with structured thinking. 

*

Dr. Melonie Richey teaches AI courses to include The Tradecraft of Data Science at federal, commercial, and academic organizations in and around Washington, DC. In the coming blog posts, she will explore knowing, learning, and expertise in the age of AI.