Problems, Teams, and Methods
In his book The Logic of Sense, Gilles Deleuze devotes a significant portion of the text to considering the concept of problems. For Deleuze, problems are a state space of n-dimensions composed of differential elements and delineated from other problems by singularities. These singularities are points where a qualitative transformation occurs, i.e. an event happens. The state space is a “problem” because the differential elements are by definition unequal to each other and thus absolutely incompatible. Any compatibility is only relative to a certain subset of dimensions, so the elements are irreducible, and any stable cluster or pattern is wholly contingent and relative.
To think about this in practice, consider a boat on the ocean. The boat is buffeted to and fro by waves. Waves, in this case, are the differential elements. The motion of the ocean is created by numerous greater and smaller forces, beyond which the wave ceases to be a wave of seawater. Such forces are the chemical bonds of the water molecules, the land of the several continents and islands that surround this ocean, the moon affecting the tide, ocean currents of greater or lesser size depending on how we scope the scene, flotsam and jetsam, flora and fauna, and more; the boat must navigate all of these in order to stay afloat and sail.
John C. Brady has written a nice article elaborating on this conception. One thing that he write about that I want to dwell on is the idea of events as crises. On his reading crises are those singular points which evoke a qualitative shift along a given dimension, e.g. the boat crashing onto the shore. These are distinguished from ordinary points where a qualitative shift doesn’t occur: smooth sailing. Those ordinary points must be understood as that relatively stable, or metastable, part of the state space.
With all that said, I want to take a look at Thai Wood’s recounting of Gary Klein’s The Strengths and Limitations of Teams for Detecting Problems. In this article, Klein studies several teams and how they act in the run-up to disruptions to their work operations. Wood sums up the good of teams quickly: since they are composed of multiple people they can content with many more dimensions and do work for longer periods of time than a single person could, and can divide work and act in parallel.
However, the team doesn’t always attain or maintain the “common ground” necessary to ward off trouble. It may not integrate warning signals from a portion of its members and thus can encounter trouble in spite of the presence of sufficient knowledge. It may also keep information siloed so that some members focus on other tasks and thus aren’t prepared to help with the impending issue. (This doesn’t just apply to human teams. Human-machine teams encounter the same dynamic.)
Furthermore, in the midst of an incident continual sensemaking is required to achieve success. Returning to the boat metaphor, it takes the whole crew to dock a ship and avoid crashing into a pier. Teams can avoid this by designating someone as Incident Commander (or captain) but then that person has outsized power and may suppress or ignore legitimate issues. Finally, teams operating in ordinary times may have rivalries or different expectations of data sources which can impede cooperation.
All this is to say that there’s a lot of factors which go into considering how teams work through problems. And I do mean problems in the Deleuzian sense. If we work with that sense, the goal of teams is to continue the problem as currently constituted along certain dimensions. These dimensions are the ones that remain in the event of load shedding, and may be different from what any one member or even all the individual members would choose to prioritize; that’s because the team is an entity which is irreducible to the members and has its own specific set of problems that it must handle.
The team is what emerges once common ground is attained, and its members’ differential relations must remain metastable in order for it to operate and to be able to ward off singular events which would destroy it or induce qualitative changes along its prized dimensions. The task of “warding off” is to be understood as keeping events ordinary, and is I think the sweet spot that brings together safety as used in “safety science,” reliability in the sense used in “site reliability engineering,” and resilience as used in “resilience engineering.”
If that is indeed the case, then I think we can propose some practical steps for aiding people who are doing that warding off. One that folks working in software have been promoting is the “Learning from Incidents” approach to Learning Reviews. (I prefer to call these “Learning Reviews,” per Ivan Pupulidy, rather than “Incident Reviews” simply because I think the latter has too much dead-weight in software. However, I recognize this is a terminological and branding decision which is at odds with many of my peers in the industry.) I take it that the act of analyzing and performing this Review is conducted by an “Incident Analyst.”
The LFI method is one I whole-heartedly endorse. Another that I want to suggest is that folks look at team behavior via formal, mathematical models. This, to me, seems like the next frontier not because qualitative studies are no longer useful but precisely because there is so much good material in that vein! Forerunners in the safety sciences didn’t rest on the canon they received: they made their own contributions. And it seems to me that many aspects of team dynamics like confidence in particular information sources, communication frequency and latency, the number and types of incidents worked together, and other dimensions could be built into something like an Agent-Based Model. Even things beyond a given team, like the diffusion of information through post-incident analyses conducted via tools like Jeli could be interesting and useful.
The great thing about bringing Deleuze into this is that he gives nice conceptual clarity to the situation teams face today, and that clarity enables us to think in new ways. It lets us understand what it means for a team to “detect problems,” explain why teams behave in the various ways they do, and then to consider how we might bring those concepts and abstractions to bear on concrete situations. We can create something new which might help us solve the new problems we face today.