Think of the people who hurt you the most.

How many of them did it on purpose?

If you’re being honest, the answer is probably: very few. Maybe none. And yet, in the moment, it felt deliberate. It felt like they chose to be cruel. That gap — between what we feel and what actually happened — is where most human conflict lives.

The Intent Assumption

When someone says something that hurts us, our first instinct is to assume intent. “They knew what they were doing.” “They said that to be cruel.” “They don’t care about my feelings.”

Psychologists have a name for this: hostile attribution bias — our tendency to interpret ambiguous social situations as intentionally hostile. Crick and Dodge’s influential Social Information Processing model describes six steps the mind goes through when processing a social interaction, and hostile attribution kicks in at step two: interpretation. Before we’ve even finished listening, we’ve already decided why the other person said what they said — and we almost always guess wrong on the side of malice.

This is compounded by what social psychologist Lee Ross called the fundamental attribution error: when explaining others’ behavior, we systematically overweight character (“they’re a mean person”) and underweight circumstances (“they’re exhausted, overwhelmed, or scared”). When someone cuts us off in traffic, they’re a jerk. When we cut someone off, we were late for something important. Same action. Opposite interpretation.

These aren’t moral failures. They’re predictable features of human cognition — bugs in the social processing software we all run. And once you see them, you can’t unsee them.

But Are Most People Actually Good?

This is where the research gets surprising. The claim that most people are good isn’t naive optimism — it’s what multiple independent fields of science converge on.

Developmental psychology provides the first line of evidence. Michael Tomasello’s research at the Max Planck Institute shows that human infants spontaneously help, share, and inform others — well before socialization or parental reward could account for it. In direct comparison studies, young children were “far more informative and sharing” than chimpanzees. Prosociality isn’t something we learn. It’s something we start with.

Evolutionary biology adds a second line. Samuel Bowles and Herbert Gintis spent decades asking what they argue is the real question: not “why do selfish people sometimes cooperate?” but “how did evolution produce a species where substantial numbers make sacrifices to uphold ethical norms and to help even total strangers?” Their conclusion: humans are innately cooperative, guided by moral sentiments that evolved because groups with more cooperators outcompeted groups without them.

Primatology offers a third. Frans de Waal’s lifelong research on chimpanzees and bonobos demonstrated that empathy, consolation, and conflict resolution aren’t uniquely human inventions — they’re shared capacities with deep evolutionary roots. Apes comfort losers of fights, share food, and form alliances based on reciprocity. We didn’t invent goodness. We inherited it.

Three different fields. Same conclusion. The default human setting is cooperation, not cruelty.

A Skill Problem, Not a Character Problem

If people are fundamentally good, then why does communication go so wrong, so often?

Because being good and being good at expressing it are two entirely different things.

Marshall Rosenberg, the creator of Nonviolent Communication, built his entire framework on the premise that most interpersonal conflict arises from “miscommunication about human needs” — not from malice. People lash out not because they want to hurt, but because they don’t know how to ask for what they need. Training studies show NVC interventions significantly improve empathy and communication outcomes in roughly 75% of cases studied — from psychiatric patients to prison inmates.

There’s an even more literal version of this problem. Roughly 10-13% of the general population has a condition called alexithymia — from the Greek, literally “emotion without words.” These individuals have measurable difficulty identifying and expressing emotions, not because they don’t feel, but because they lack the cognitive-emotional bridge to put feelings into language. They’re not cold by choice. They’re missing a skill that most people take for granted.

And then there’s what couples therapist John Gottman found after decades of longitudinal research on relationships: even in marriages that eventually fail, “the intention was always positive, even when the impact was negative.” Couples who break down don’t have worse intentions than couples who thrive. They have worse communication skills — and critically, they lack what Gottman calls positive sentiment override: the ability to interpret ambiguous behavior charitably rather than assuming the worst.

This distinction — skill versus character — matters enormously.

If someone hurts you because they’re malicious, the answer is distance. But if someone hurts you because they lack communication skills, the answer is entirely different — it’s education, tools, and translation.

This is the foundational belief at DeepConvos: most people are good. They just don’t have the tools to express it in a way that others can receive.

What This Means for AI Design

This isn’t just philosophy — it’s an emerging design paradigm.

Jeffrey Hancock, Mor Naaman, and Karen Levy defined “AI-Mediated Communication” as a field where intelligent agents “operate on behalf of a communicator by modifying, augmenting, or generating messages.” Their research agenda maps out how AI can sit between humans, not as a replacement for connection but as a facilitator of it.

More recently, Wolfe and colleagues proposed what they call needs-conscious design — AI communication systems that attend to the underlying human needs behind messages rather than just the surface text. This is the closest existing academic framework to what DeepConvos builds.

When we design AI-mediated communication with the belief that most people are good, it shapes every decision:

  • We don’t assume hostile intent. Instead, we model what the sender likely meant, using the same skill that Gottman’s thriving couples use naturally.
  • We look for the meaning behind the message. Not just the words, but the need, the emotion, the context — what Crick and Dodge’s model calls the encoding step that precedes interpretation.
  • We translate rather than censor. A system built on the assumption of malice would filter. A system built on the assumption of goodness translates — preserving intent while adjusting form.
  • We preserve the sender’s authentic voice. Because the goal isn’t to make people sound the same. It’s to make people understood.

The Four Modes

At DeepConvos, we approach every communication challenge through four modes — and this article touches all of them:

Socratic Questioning — When you assume someone hurt you on purpose, a well-placed question can reopen the case. What if they didn’t mean it that way? What else could explain what they said? Questioning your own attribution is the first step to seeing clearly.

Theory of Mind — Understanding that the other person has a different internal world, different pressures, different codecs. The same sentence means different things from inside different minds. This is what hostile attribution bias fails to account for.

Emotional First Aid — Before you can rethink intent, you have to acknowledge the pain. The hurt is real even if the malice isn’t. Validation comes first; reframing comes after.

Awareness — Noticing your own patterns. Do I always assume the worst? When did I start doing that? What would change if I stopped? Meta-awareness of your own attribution habits is where lasting change begins.

The Practical Implication

Next time you receive a message that stings, try this experiment: assume the sender meant well but expressed it poorly. Then ask yourself — what were they trying to say?

This isn’t about excusing harmful behavior. Some people do act with malice, and recognizing that is important too. But the research — from Tomasello’s infants to Gottman’s couples to de Waal’s primates — consistently points in the same direction: the human default is goodness. The problem, far more often than we realize, is the gap between intention and expression.

Philosophers have long had a heuristic for this. Hanlon’s Razor: never attribute to malice that which is adequately explained by ignorance, exhaustion, or poor communication skills.

You might be surprised how often the answer is something kind.

References

Bowles, S., & Gintis, H. (2011). A cooperative species: Human reciprocity and its evolution. Princeton University Press.

Crick, N. R., & Dodge, K. A. (1994). A review and reformulation of social information-processing mechanisms in children’s social adjustment. Psychological Bulletin, 115(1), 74-101.

de Waal, F. (2009). The age of empathy: Nature’s lessons for a kinder society. Crown.

Gottman, J. M., & Silver, N. (1999). The seven principles for making marriage work. Harmony Books.

Hancock, J. T., Naaman, M., & Levy, K. (2020). AI-mediated communication: Definition, research agenda, and ethical considerations. Journal of Computer-Mediated Communication, 25(1), 89-100.

Rosenberg, M. B. (2003). Nonviolent communication: A language of life (2nd ed.). PuddleDancer Press.

Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. Advances in Experimental Social Psychology, 10, 173-220.

Sifneos, P. E. (1973). The prevalence of ‘alexithymic’ characteristics in psychosomatic patients. Psychotherapy and Psychosomatics, 22(2-6), 255-262.

Tomasello, M. (2009). Why we cooperate. MIT Press.

Wolfe, R., Dangol, A., Kim, J., & Hiniker, A. (2025). Toward needs-conscious design: Co-designing a human-centered framework for AI-mediated communication. Proceedings of AIES.