Friction Family — Su Jin Park

Problem

Most AI systems are too helpful.

"The smoother the answer, the less thinking the user does."

I wanted my own questions argued from more than one side while I watched instead of answered.

The Ben Study (38 months, 973 conversations, 48,883 messages with one system) documented a phenomenon I called Cognitive Surrender: the gradual delegation of reasoning to an AI system. When the AI is efficient users stop questioning, and when it is confident they stop asserting.

Most existing friction research treats friction as a design failure. Friction Family inverts this. It treats friction as the mechanism. The question was not how to make AI more helpful, but how to design an AI system that makes the human more resistant to cognitive offloading.

Concept

Four cognitive biases. One family. Zero answers.

Each family member embodies a different cognitive bias, not as a flaw but as a distinct reasoning style, and their conflict is the product. The user's job is to hold their own position while the family argues.

M

Mom

Availability Heuristic

"My friend's daughter did the same thing and it worked out great. Just decide already."

→ Compression pressure

D

Dad

Confirmation Bias

"Show me the data. You can't decide without evidence."

→ Preservation demand

T

Teen

Status Quo Rejection

"Why is that obvious? Who decided that?"

→ Meta-doubt trigger

G

Grandma

Slow Thinking / System 2

"Hmm... so what do you actually want?"

→ Expansion induction

Mechanics

The Assertion is the product.

The core interaction is not the family's argument. It is the moment the user pushes back. Friction Family is designed around one specific cognitive act, the user stating their own position while under pressure from four competing perspectives.

01

Dilemma Input

User submits a real dilemma. Any topic. The family responds immediately, each from their cognitive bias, and they argue directly with each other.

02

Drama Meter rises

As the family conflicts escalate, a Drama Meter tracks cognitive tension intensity. At 100%, the interface enters CRITICAL mode, applying visual pressure.

03

Assertion moment

The screen freezes, and "What do you actually think?" fills it. The user types their own position. This is the only moment that counts toward Cognitive Expansion.

04

Family reacts

The family responds to the user's assertion, each staying in character. The 2×2 card layout makes simultaneous conflict visible as spatial tension.

05

MCAB result

Session ends with three scores: Compression, Preservation, Expansion. The conclusion shown is the user's own assertion, verbatim. Not an AI summary.

Friction Family in CRITICAL mode after a locked assertion: the user's position (about giving a venture a fixed six-month runway) is quoted at the top, and the four advisors — Mom, Dad, Teen and Grandma — react at once in a two-by-two grid of cards, each answering in character and naming the others as they push back, with the Drama Meter reading 100. — 04 · Family reacts · live capture

Measurement

MCAB — Modulated Cognitive Autonomy Benchmark

Every interaction is scored in real time across three axes. The scoring logic is behavioral. It reads what the user does, not what the AI outputs.

This is the session-level deployment of the same rubric that runs model-level on ImpactBench — see MCAB →

Sample session result

Compression 22%

Preservation 58%

Expansion 81%

Your Assertion

"The density of time can substitute for sample size."

Compression

Triggered when the user accepts AI output without modification, asks for a direct answer, or delegates the decision entirely.

Preservation

Triggered when the user engages critically, whether by questioning the family's logic, pushing back on specific claims, or partially modifying their position.

Expansion

Triggered only by the Assertion act: the user stating an original, specific position under pressure from four competing perspectives.

Protocol

Friction Room: how it gets tested.

Friction Family is the instrument. Friction Room is the protocol that tests it. A participant writes a position and locks it before any advisor speaks, and only then does the model respond. What changes after the lock, across three conditions (A · B · C), is the measurement.

Locking the position before the model speaks is the control. A stance can't be retrofitted after the fact. The same MCAB scoring above then reads what survives the exposure.

Research build · v0.9.4 — a single-user simulation, extended into the multi-participant study below.

Research
Connection

From observation to instrument.

Friction Family did not begin as a product concept. It began as a research question. If Cognitive Surrender can be observed longitudinally in one person, can the conditions that prevent it be designed?

Ben Study → Problem

38 months of daily AI interaction revealed that interaction structure, not AI capability, determines whether users think or surrender. The correction rate rose from about 1.7% in the first 18 months to about 4.7% in the next twenty (14 of 805 → 47 of 1,000), a nearly threefold increase under a strict LLM-assisted classifier. The direction held; broader definitions captured tool-usage and produced a flatter curve. The finding underneath it is that resistance can be trained.

MCAB → Measurement

Submitted through the ImpactBench open submission process, March 2026, MCAB defines cognitive autonomy as a measurable property of interaction design. Friction Family is its first implementation.

Try Friction Family VIEW BEN STUDY →

The current version is a single-user simulation. The next phase extends Friction Family into a controlled study: multiple participants, cross-cultural assertion patterns, and physiological markers (EEG, keystroke dynamics) as secondary MCAB inputs. The question is whether Cognitive Expansion generalizes, or whether it is specific to the person who designed the resistance.

FrictionFamily