Turning weekly work lessons into academic wisdom.
There are dozens of organizations that design operational infrastructure systems, both for internal use and for selling solutions externally. A peculiar characteristic of any such system, whether it’s a power plant control room or a military command center, is that most, if not all, of them feel intimidating. That intimidation arises from the inability to fully comprehend the unfamiliar. Without comprehension, understanding becomes impossible and that is one of the most intimidating things imaginable. It’s why we fear the dark and growing old.
I particularly enjoy observing newcomers like The Nuclear Company, Merlin Labs, and Interlune: they serve as a reminder that the commodification of thought only applies to problems largely solved and that there are still challenges beyond those solved by the zeitgeist’s “vibe-solutions”.
The “Let’s Punt On It” Problem
The fundamental challenge of any intricate operational system designed to make timely and precise decisions lies in the inability to gain situational awareness (SA) solely by collecting all available data.In the early days of software development, a reasonable approach to designing such systems was to present all available information to users and just let them make their decisions on their own. The systems addressed problems that were so complex that delegating the high-stakes decision-making to end users seemed like the most user-friendly approach. However, as software became increasingly intuitive and beautiful, legacy operational infrastructure systems remained stuck in the past. Therefore, it is worth exploring what makes them so difficult to redesign and how we can overcome it.
The Solution. Goal-Directed Task Analysis
Every intricate system at some point needs to present data to a human responsible for assessing, recognizing, classifying, and comprehending their current situation in order to take the right action. While number of actions a human can perform is limited, the data available for analysis is, for the purpose of this argument, limitless.This persistent problem of information overload necessitates a framework that prioritizes the goals of operators to mitigate significant cognitive bottlenecks. The problem is often described as drinking out of a firehose, suggesting that although you could attempt to consume a powerful stream of water meant to put out fires, it isn’t in your best interest to do so. But if your goal is to drink, using a hose in most situations wouldn’t be your first choice.
To make a decision, people first need to understand and categorize the situation they’re in, otherwise known as gaining situational awareness (SA) – is there fire or am I just thirsty. Recognizing familiar situations is crucial for operational effectiveness. In fact, when faced with operational pressure, decision-makers spend most of their time assessing and classifying the situation, leaving just enough time for executing. Once the SA is established, a course of action we’ve learned or encountered before is immediately adopted. And since it is the potential action that determines which data needs to be acquired for assessment, we’re no longer dealing with limitless data. Therefore the optimal approach for a designer is to translate a finite number of actions a user can take into a handful of situations with a limited number of first degree data available that is relevant. This approach reduces the number of data points worth surfacing, in turn accelerating the process of gaining SA from reading it. Once the situation is recognized, the path to action is, again, relatively straightforward.
Acquire the relevant data. Gain situational awareness. Predict the future.So in order to overcome the problem of information overload, we must first remove the data irrelevant to the decision, if at all possible. Recognition being a key factor, you wouldn’t send a junior to fix anything they haven’t seen before: pattern-matching and drawing upon long-term memory structures to arrive at a solution is something only veterans can do. Consider this for your onboarding and tutorial flows.
Designing (or redesigning) a system where situational awareness acquisition is the primary key performance indicator (KPI) of the interface means:
- providing data that supports user tasks in the order of their use
- designing systems that facilitate operators’ ability to acquire and maintain SA in dynamic environments
- structuring dynamic visual representations around operator goals to support the development of high levels of SA across the UX
- synthesizing multiple telemetry points into higher-order states to reduce the mental workload of reviewing every piece of data
The Fallacy
However, a pitfall many designers fall into is translating raw data into synthesized system messages one step removed and dismissing the raw points altogether. Although slower in arriving at the right decision, a system designed for showing just the “dumb” data points will outperform a system showing the synthesized right answer.This problem, long thought of as solved, has recently resurfaced with the plethora of AI tools and AI-powered software that is promising to make the right decision for the decision-maker. But this begs for a separate discussion altogether.
E.g. Google Maps “11 minutes away” message might answer the question of ETA faster, but for the human assessing the situation – are they late, do they have enough gas, is there time for another episode of their favorite podcast – the raw “10.3 miles” message is just as useful. They are in a sub-process of gaining sub-SA and the derived data point of 11 minutes could become a cognitive bottleneck: “are we just 11 minutes away because I’ve been speeding for the last five minutes, or is there no usual traffic on the way?”.
Humans spend most of their time understanding and classifying situations when making decisions, and only a small fraction of time on acting, whether it’s deciding to inform someone they’ll be late or slowing down to enjoy their favorite show in the right lane. So as designers and human factors engineers, let’s not add to their mental effort of understanding the situation.
It’s worth noting that, given that Google Maps serve many intents, and are used to answer both the “when” and the “where” questions, the UI, unable to accurately predict the specific intent, surfaces both the raw and the derived data simultaneously.
The Solution
To recap, in order to support your user’s quick acquiring of SA with all the data captured and ready to be surfaced, with the end goal of making the right decision quickly, it is advised, to:- Figure out the limited number of situations your users will experience while interacting with your software
- Outline the minimum raw data required for gaining situational awareness within the said sutiations based on all possible user goals and intents
- Present the raw data in a manner that makes it easy to be acquired and impossible to miss
- Attempt to surface derived data points in a secondary manner, in a way that supports SA acquisition and externally confirms (or clearly disagrees with) the decision made