The Number Whisperer
My wife’s pretty direct. She’ll say it as it is, whether you like it or not. As a high-school math teacher, this comes in quite handy. Math isn’t poetry, not in high-school anyway! Being direct is a very efficient way of scaling communication with hundreds of students. Sometimes kids get offended with this directness. They cry foul at home, the parents then cry foul to the principal. On one such occasion, the principal patiently listened to the parent’s complaint and then said in defense of the teacher: “Don’t mistake certainty for aggression.”
Context Windows
When it’s time to eat, my dog looks at me a certain way. There’s yet another way he looks when it’s time for a walk. A slightly different look for when he’s ready to play. All nonverbal cues and all clearly understood by the family. If there was ever a clearer demonstration of intelligent, nonverbal communication just with the eyes, I don’t know what it’d be.
My dog’s smarts don’t need a vocabulary
So what does speech add? If you believe Robin Dunbar, it’s social bonding. The hypothesis is that primates used physical grooming to form social bonds. Language was the evolutionary hack to scale beyond small primate groups. Early language might well have been just tones and volume but it didn’t replace the underlying social cohesion, it just layered in another dimension.
As human beings, we all carry slightly different versions of that underlying social fabric. When communicating with other human beings, if we forget about the hidden layer, our words don’t mean what we want them to. Two people can use the exact same words and walk away from two completely different conversations.
And because everyone carries a slightly different version of that social context, everyone filters words differently. Some people are raised to hear tone as the whole message. Others are raised to take words literally and let tone pass through. That’s just one part of the filter; there are so many factors that influence our filters: gender, volume and hierarchy are some that come to mind. It’s not hard to see why communication with language is such a darn hard problem. Nobody’s “wrong”. They just have different filters and when the filters don’t match, a random dominant factor wins.
Gender
How often do you subconsciously label a talkative woman “aggressive”? And how often do you label reserved women as “incompetent”? How often do you do this for your own family, friends or work circles? You know you do. Women often face a double bind. Speak up to match the male voices in the room or soften your stance. Heads I win, tails you lose.
In a study of U.S. senators, Yale researchers found that power increases talking time for men: the more senior the senator, the more floor time they took. For women, there was no relationship between seniority and speech. Powerful women self-regulated and constrained their speaking, staying quieter than their male counterparts. Women adjust artificially to compensate for the double bind, but this only leads to both genders penalizing women equally.
How did we develop the gender filter? Turns out that a lower timbre registers as dominant across just about every species that vocalizes, ours included. That in turn happens to be one of the more reliable things puberty does to a male body. It sucks, but it just means that the oldest, crudest gauge we have for “who’s in charge” was calibrated on pitch. That evolutionary expectation creates the bias and penalizes women who aren’t in line with the expectation.
Heads I win, tails you lose
Volume
Feel like you’re not being understood? Feel like the gender bias is getting to you? Feel a bit insecure? There’s a simple hack - speak up! Volume as a substitute for substance feels like authority. Similar to timbre, just less likely to suffer from the gender bias. No guarantees on whether that gets rewarded or penalized though.
Your audience decides if your volume will be perceived as certainty or a performative act that’s compensating for something. Does your content have substance or is the force of delivery doing the work? As it stands, our brains are wired to pay attention to louder voices regardless of what those voices are saying.
Evolution sucks…again. Biologically, high acoustic intensity projects dominance across both male and female speakers, likely as a sustained threat perception mechanism. The layers are still intertwined though; the same mechanism deteriorates significantly when filtered through social, organizational or gender biases. What’s more, relying on loudness as an authoritative cue can backfire; high volume disorients listeners, provokes amygdala-driven threat responses, and escalates conflict, ultimately turning intended authority into defensive shutdown.
Still…scream.
Hierarchy
I’ve had many moments of late when I’m couching my words at work. I’m sure this plays out for you too. Your boss says something you know is wrong, but you choose your words carefully. There’s a tradeoff at play and conflict avoidance almost always wins.
Evolution…again. Animal pecking orders established a clear chain of deference to reduce energy-draining physical fights over resources. In modern organizations, titles and seniority function as formal social proxies for the same primitive biological pecking order.
For the most part, titles and seniority carry enough ego baggage that couched communication up the chain doesn’t matter; like I said, conflict avoidance is a worthy tradeoff. However, sometimes, using rank and titles as a proxy for absolute truth can be dangerous. Remember the string of failures at Boeing? Or flawed cockpit communications like the ‘97 Korean Air disaster?
Loud Language Modeling
Our use of language and nuanced communication demonstrates intelligence unique to humans. But language is just the scaling layer on top of an intricate social fabric; no surprise that human beings can say so much with gesture, tone, presence and facial expressions alone. Unfortunately, it’s a world where the very non-verbal nuances that allow us to build trust, extend empathy, and soften conflict are the same biological channels that trigger stereotyping, status hierarchy, and vocal bias. The discussion here barely scratches the surface on these layers and biases.
Modern LLMs have no way of encoding non-verbal, unspoken or unwritten nuance. So while the best trained models can produce perfect language, structure and grammar, they risk retaining all the flaws, stereotypes and biases human beings have created over generations.
We’re entering an era likely to be dominated by synthetic communication. AI removes the biological traps of tone, volume, and gender, but at the cost of the social context that holds human relationships together. Human intelligence sits between tempering our own instincts and maintaining a healthy distrust of the generated world.
Rabbit Holes
Robin Dunbar, Grooming, Gossip, and the Evolution of Language (Harvard University Press, 1996). Language as scaling for primate grooming.
Victoria L. Brescoll, Who Takes the Floor and Why: Gender, Power, and Volubility in Organizations (Administrative Science Quarterly, 2012).
John J. Ohala, An Ethological Perspective on Common Cross-Language Utilization of F0 of Voice (Phonetica, 1984). Lower fundamental frequency reads as larger and more dominant across most vocalizing species.
John G. Neuhoff, Perceptual Bias for Rising Tones (Nature, 1998). A sound getting louder is consistently judged as more urgent and closer than an equivalent drop in volume, regardless of who or what is making it.
Nicolas Burra, Leonardo Ceravolo, and Didier Grandjean, Early Spatial Attention Deployment Toward and Away From Aggressive Voices (Social Cognitive and Affective Neuroscience, 2019). An angry-toned voice captures and holds attention almost twice as long as a joyful one, measured down to the millisecond by EEG.
John Maynard Smith and G.R. Price, The Logic of Animal Conflict (Nature, 1973). Why animals settle into ritualized dominance hierarchies rather than escalating every contest to a real fight.
NTSB, Aircraft Accident Report: Korean Air Flight 801 (2000). Cockpit communication was one of several contributing causes.
House Committee on Transportation and Infrastructure, Final Committee Report on the Design, Development, and Certification of the Boeing 737 MAX (2020). An 18-month investigation into the two 737 MAX crashes (346 dead) found production pressure led engineers to conceal a flight-control system, MCAS, from the pilots who needed to know it existed. Rory Kennedy’s Downfall: The Case Against Boeing (Netflix, 2022) covers the same ground and is worth watching.