FIGURE DRAWING: OBSERVE AND REFINE

Relating ethnography and data science practices begins in the studio of an artist with a metaphor; figure drawing. An exercise that requires a clothed or nude person (most common), figure drawing is a traditional fine arts practice used by an artist throughout his/her career to continually develop foundational drawing skills. More importantly, figure drawing is an exercise about refinement of observation.

An artist drawing a model only glances at their sheet to mark an observation during their session; most of the artist’s time is spent observing the model. Drawing is a process of refinement: observe, refine/mark; repeat. The goal is not to render the image in a single pass; rather it is to develop the image over a defined period of time. An artist’s gaze fixed too long on the drawing as he/she draws, reveals a practice that results in an image describing what the artist imagines the subject to be rather than how he/she exists.

Evidence of a drawing’s development (i.e. observations) can be found within the drawn artifact. Each of the figures in Figure 1. and Figure 2. describes a unique pose by a single model that was held over a specific duration of time (i.e. “5 minute pose”). Figure 1’s knee is formed over the course of at least 3-4 observations; progression is most visible with this figure in by the number of lines required by the artist to define the knee. Early marks are visible that were refined over time. Development of a drawing is more visible in Figure 2. With the number of marks defining each part of the human form layered over each other.

Observe, refine/mark; repeat. In the context of building and iterating through repeated observations, an artist’s practice is similar to that of an ethnographer and data scientist. Not all marks are correct, but they create a set of knowledge that is directional, leading to an accurate solution. All three practices begin with observations and broad marks that are iterated on over time, resulting in a meaningful artifact that is revealed slowly. [1] Images by Tony Cheng

Figure 1. Figure Drawing

Figure 2. Figure Drawing (ran out of time)

INTRODUCTION

The author of this paper is user experience researcher at Uptake, a company specializing in artificial intelligence in the space of Industrial IoT. Within Uptake, software development leverages core resources throughout an engagement, including data science and user experience research. Due to this ongoing working relationship with the data science team, Uptake’s user experience researchers have gained a unique understanding of data scientists and their everyday practices.

This view provides Uptake’s UX research team an empathetic understanding of and appreciation for the manual, tedious and often-invisible process occupying most of a data scientist’s time. Their effort begins with a problem defined by a client, one self-identified or surfaced through field research. Then an ongoing cycle of observing and refining their subject by identifying a source(s) for the required machine data; generating and logging that data, engaging with it, confirming that it is the correct data, and ultimately crafting purposeful models to yield actionable insights.

Data—quantitative and qualitative, supporting data science or UX—in its raw forms is an observation, not evidence. A machine’s gears are grinding, the temperature reading is 400° F, a bus will not start, an asset’s fluids levels are x, a component is on/off, etc. A single observation is just as a mark on a sheet a paper that contributes to the rendering of a figure.

Actionable insights emerge from this data when rich contextual and historical data is introduced. For example, an alert for a rail engine’s loss of horsepower will be triggered when a certain threshold is met. Observation.

This alert is observed in the data during each run at the same location along a specific route. Does the engine need to be serviced? No. Evidence is forged from quantitative and qualitative observations. Upon further inspection, the train passes through a tunnel that has limited circulation, forcing the engine’s intake center to be temporarily overwhelmed by the exhaust system, triggering an alert for loss of horsepower. Horsepower returns to normal levels immediately after exiting the tunnel.

Raw sensor readings coupled with qualitative context becomes evidence supporting—in the example above—a decision not to service the engine and ignore all alerts of this type generated at this location. Action may be taken if the same fault occurs at a different location along the same route.

Observe, refine/mark; repeat.

Industrial IoT’s soul is large data sets and real world context that require technical excellence and human curiosity to generate meaning from it through an ongoing cycle of observing, refining and marking.

UPTAKE DATA SCIENTISTS AND USER EXPERIENCE RESEARCHERS

A typical figure drawing session is organized as a ring of people, each with a pad of paper and a drawing instrument around a small rise; at the center of this group is a single model posing on the rise. At the conclusion of the pose, each person has rendered an image of the same model from a unique view.

Individuals drawing during these sessions are usually fixed to a bench or easel for the duration of the session as a matter of convenience. Models will change poses and their orientation to the group throughout a session, allowing each person to “see” different views of the same model. An alternative approach is to have a model hold a long pose as a group of artists (or single individual) rotate around the model in timed increments to see the same subject from multiple points of view.

A drawing session is an effort to understand a figure through individual poses and views of the model. An artist restricting his or her a view of the model to only front, side or back poses limits their understanding of how the complete form exists in space.

Poses can last for durations of 30 seconds, 1 minute, 5 minutes, etc. Or a pose may be continuous for hours across multiple sessions. No matter the length of time, the initial marks made a on sheet of paper situate the figure within the page’s space. These marks do not commit the artist to a final image, that image develops from continuous observation of the model and piecemeal refinement of the drawing.

A project begins at Uptake with each role positioned around the project’s challenge. Uptake’s data scientists need access to large sets of data—generated by man and machine. Data is not always received in an organized and tidy package ready to be acted on. Data scientists, working with subject matter experts (SME), begin their effort by understanding the data available to them. (Including the asset(s) generating data.) From this initial view, they can determine what additional data and quantity of it is needed. If the required data does not exist, they develop a plan to generate and log it.

Observe, refine/mark; repeat.

As data scientists engage machine data, user experience researchers are understanding the context surrounding the data through preliminary research and SME engagement. This is critical learning ahead of any fieldwork due to the nature of Industrial IoT sites. They are often dangerous spaces with unique obstacles, including: intense security/regulatory requirements (Department of Defense, rail), remote locations that are costly to access (mining), limited number of candidates to speak with (manufacturing), and a required escort for the duration of a visit (all).

Observe, refine/mark; repeat.

Similar to the initial broad strokes of an early figure drawing, these first steps of discovery are to situate the roles before committing to courses of action. Data sources and end users may change over time just as the figure’s final image emerges through a series of observations and marks.