Throughout human history, it has been important to be able to make quick decisions, especially if the outcome of a wrong decision might lead to injury or death.1 Consequently, our brains are pre-wired to be incredibly adept at generating predictive algorithms and generalizations from previous inputs (experiences), even if the data is incomplete.2 This is a legacy issue with our brains – that we have a natural tendency to extrapolate signals, even from random noise.
This lesson was most acute to me while conducting my research for my Master’s degree in acoustical engineering at Penn State.
Mistaking Ambient Noise for Signal
As part of my thesis, I was to determine experimentally, the flow noise produced from a spherical sensor as it moves through fluids of varying viscosity.3 Sparing you the details as to why this was even research, suffice it to say, the Navy was interested in this topic and how it relates to underwater listening devices scattered around the oceans.
Numerous times, I thought I had collected good data and had completed the experiment. “I can finally start writing my thesis.”, said I… only to find out the data was just noise (the unwanted kind as opposed to the noise/signal I was trying to measure). In other words, I was looking for flow noise from the sphere, but I was measuring ambient noise in the room.4 The signal was there, just buried.
Mistaking One Signal for Another
One night, the experimental data showed an interesting spike. As the velocity of the spherical sensor increased, the frequency of this spike in the data moved up proportionally. This could be an interesting find. I was on to something. Maybe something no one has previously measured. Exciting times in the lab indeed! I exported the data and printed graphs for the team meeting the next morning.
Turns out, the monofilament line5 used to suspend the sphere in the fluid wavered as it passed through the fluid, creating trace vibrations that my acoustical sensor picked up. Basically, I was measuring noise from the test apparatus, not the noise I wanted to measure from the sphere.6
Rather than confusing ambient noise for signal, this was a case of confusing an actual signal of something uninteresting for the fainter signal I wanted to measure.
So, I continued…
Experiment >> Gather Data >> Contemplate >> Incremental Improvement. Iterate.7
Sometimes real correlations are obscured because they are masked beneath the ambient noise of everyday life. This produces ignorance (even if only in the mild form) and potentially missed opportunities.8
Sometimes we confuse noise for signal, falsely correlating random noise with a dependent variable. This is a faulty causation problem. We often fabricate a signal from the noise and assume a near-term event is correlated, establishing a cause-and-effect relationship where there is none. This allows us to assign causality from randomness. It’s also where faulty thinking and truly weird ideas come from. Things like: “Every time I wear my rubber chicken hat, my favorite team wins their baseball game.” If we are to have better mental models of the world, sometimes we need to dig below the surface to get the real signal and to establish the correct cause-and-effect relationships.
Sometimes we confuse one signal for another. We do this in our experimental data in the science lab, with our business data, and with the data in our lives. Because we are so predisposed to narrate a story to fit our observations, we are prone to create entirely false narratives to fit the wrong signal.
The problem with these errors in our thinking lies in the stories we tell ourselves – completely fabricated stories contorted to fit our observed data. These stories represent our model of the world for the data we have seen. But they may have very little predictive value.
Worse, we may have unfounded levels of confidence in our interpretations of how the component parts relate to each other and our predictions of how they will work together in the future. This leads to poor decisions in the now. It’s one thing to be wrong and acknowledge we might be wrong. It’s another to be wrong with confidence.
Questioning Our Models
We should always question our mental models and world views and be open to the idea that we could be wrong. We might be wrong due to lack of signal (equivalent to having no prior experience or exposure to a topic or concept) or due to false signals (incorrectly interpreting prior data).
I plan to write considerably more on the concept of being wrong, frameworks of thinking and the need to continually question our foundational assumptions. More to come on these topics…
P.S. Follow Past Midway if you would like an email notification of new posts.
- People often write “leading to injury, or worse”… but that sounds even more ominous. Besides, isn’t there only one option “worse”?
- I originally wrote “even if the data is scarce”. But data is never scarce. Through our senses, the brain receives a continuous flow of tremendous quantities of data. The idea that we can filter through the noise and keep only the most important parts for decision making is quite incredible when you think about it.
- Exciting, I know.
- After discovering this, I had to perform my experiments after midnight when the building ventilation shut off and the human commotion stopped (traffic, people shutting doors, footsteps in the building, etc.). It took two people to perform the experiment. God bless my wife for all those nights up at the lab with me past 2am. She’s a saint.
- “monofilament line” sounds way more technical than “fishing line” and is thus more suitable language for a thesis in sciencey-stuff.
- A fellow grad student who was a few years older than me (with actual acoustical work experience) pointed out that it was likely the suspension line. A quick calculation of the Strouhal number showed it was almost certainly the line because the vibrations matched the theory for vibrations of a very small cylinder, not a larger sphere. Bummer. Fortunately for me, I had smart colleagues with some experience that kept me from drawing false conclusions “or worse”, reporting it in my thesis.
- For my research, this process seemed like an infinite loop. I thought I would never finish. But persistence eventually teased the data out of the noise with my experiment and I completed my thesis. Tenacity pays.
- There are many stories of love lost and romance forlorn born from weak signals masked by the excessive noise surrounding our overly-busy lives.