Primitive Asymmetries of Persistence

Primitive Asymmetries of Persistence (v1.0)

A Companion Explainer for Humans, Not Robots

1) Why This Paper Matters

Have you ever noticed people—even smart, careful ones—doing things online that seem obviously risky? Clicking, posting, or sharing in ways that can backfire later, even when the immediate benefit is tiny? Most explanations point to “bad judgment,” addiction, or manipulation—but these don’t hold up when highly informed people keep making the same choices.

This paper argues that the problem isn’t people. It’s the structure of our digital world. Online systems remember everything forever, perfectly and without context, while humans naturally focus on the present. That mismatch creates persistent risks that feel invisible until it’s too late.

2) What the Paper Says (Plain English Version)

We break the problem down into three “primitive” patterns that, together, explain why we behave this way online:

Time Mismatch (Temporal Asymmetry) – We feel rewards immediately, but negative consequences show up later, sometimes in unpredictable ways.
Records Without Context (Persistence Without Perceptual Decay) – Digital systems keep perfect records tied to your identity, but they don’t preserve the original meaning. Future viewers may interpret your actions differently.
Local vs. Global Thinking (Micro Contextual Coherence vs. Global Invariance) – We optimize for how things look right now, but the system preserves them forever, where they can be reinterpreted under different norms or rules.

These three factors flip the usual “reward system.” Our brains chase instant satisfaction, while digital infrastructure silently accumulates long-term exposure. Importantly, this isn’t because anyone is malicious—the system and humans are just operating in incompatible ways.

Where it shows up:

Social media posts and interactions
Workplace tools and performance tracking
Credentialing and government portals
Personal analytics apps

In short: most actions feel harmless—but risks quietly accumulate and can hit hard later.

3) What Makes This Approach Different

It’s about structure, not psychology. The paper says the problem comes from system architecture, not addiction, laziness, or bad intentions.
Minimal but complete. Just three core primitives are enough to explain compulsive behavior, misjudgments, and sudden panic when consequences appear. Remove one, and the pattern breaks.
Phase-based risk view. Risks don’t look dangerous at first—they “flip” suddenly under recontextualization, explaining why bad outcomes seem surprising.

4) Key Insights (If You Accept the Model)

Reward inversion is the culprit. Immediate human satisfaction is rewarded, while long-term exposure goes unnoticed. Locally rational actions can create globally messy results.
Knowing isn’t enough. Understanding risk in the abstract doesn’t prevent mistakes if the decision is made in the “local frame” without immediate penalty.
No simple fix. True alignment requires changing the system itself—adding artificial decay, context restrictions, or friction. Otherwise, miscalibration is inevitable.
It goes beyond social media. The same principles explain anxiety or over-monitoring in work, finance, healthcare, or government systems.

5) Why This Matters in the Real World

Policy & governance: Focusing on moderation or content misses the root cause. Reducing big failures requires structural fixes—like limiting how long records persist or tying them to context.
Product design: Warnings or pop-ups aren’t enough. Designers need to make long-term consequences felt at decision time or reduce cross-context record reuse.
AI alignment: Agents with memory tied to users face the same risks—old actions can be misinterpreted, causing unexpected harm. Context and decay must be built in.
Corporate risk: Old records can suddenly trigger crises if norms change. Organizations should track latent exposure, not just visible incidents.

Bottom line: The paper shows that many “bad decisions” are really the predictable result of humans operating in digital systems that never forget. Fixing this isn’t about telling people to be smarter—it’s about designing systems that respect the mismatch between human present-focus and digital permanence.