Interactive Multimodal Visualization: Helping Ordinary People Understand Machine Learning Functions

Overview

This forum post on zhichai.net introduces and comments on a research paper:

> Paper: Towards Interactive Multimodal Representation of ML Functions for Human Understanding of ML > Authors: Bokang Wang, Yingxuan Liao, Leah Lee, Jack Wesson, Anlan Yang, Ruizi Wang, Yigang Wen > arXiv: 2605.00357 | 2026-04-29

The Science Communication Problem

The post opens with a familiar scenario: a parent asks "what is machine learning?" and gets lost immediately once the answer involves "training a function" and "mathematical mapping." The author identifies core issues:

• AI/ML is too abstract for the general public
• It is saturated with jargon and mathematics, which feels intimidating
• This breeds misunderstanding and distrust — the "AI is a black box" and "AI is uncontrollable" narratives

What is needed: making ML understandable through visual, interactive, multimodal approaches that lower the entry barrier.

The Proposed Approach: Interactive Multimodal Representation

Core idea: transform abstract ML functions into perceivable, explorable forms via interactive visualization, sparking curiosity and creating a virtuous cycle of "understand → explore → understand more."

The technical proposal has four pillars:

1. Multimodal representation — not just text, but graphics, animation, possibly sound and haptics for multi-sensory understanding 2. Interactive exploration — not passive viewing, but active engagement: adjusting parameters, seeing real-time changes, asking "what if?" 3. ML function visualization — making functions visible: decision boundaries, loss surfaces, gradient flows, attention weights 4. Lowering barriers — no math or coding background required; moving from intimidation to curiosity, from fear to wonder

The author offers an analogy: traditional ML teaching is like reading formulas, while interactive visualization is like playing an "AI simulator" — tweak the knobs, watch the results, and understanding follows naturally.

Why Interactive Visualization Works

Limitations of the traditional approach:

• Abstract: formulas lack intuitive grounding
• Passive: lectures and reading invite disengagement and shallow understanding
• High barrier: math and programming prerequisites exclude ordinary people

Advantages of interactive visualization:

• Intuitive: see the "shape" of a function and understand its behavior
• Active: hands-on trial-and-error learning leads to deeper engagement
• Low barrier: anyone can participate, promoting AI literacy for all

A Feynman-Style Verdict

Quoting Feynman: "If you can't explain it simply, you don't understand it well enough." The author extends this to AI education:

> "If a machine learning function cannot be understood by ordinary people, that is not the fault of ordinary people — it is a problem with our representation. Interactive visualization is not 'simplification' but 'translation' — translating machine language into human language."

This reflects the democratization of education: knowledge should be accessible to everyone, not an elite privilege, with visualization as the bridge.

Takeaways

For anyone working on AI popularization or education, the post suggests asking:

1. Is my AI system understandable to ordinary people? 2. Can visualization help build intuition? 3. Can interactivity improve engagement? 4. How can more people come to understand and trust AI?

Core message: AI popularization requires not only technical progress but also "translation" — turning the abstract into the intuitive. When ML functions become visual tools people can play with, AI shifts from "black box" to "transparent partner." In a democratized AI future, the best technology is not the most advanced, but the most understood.

> On the bridge of understanding, visualization is the strongest brick.