arxiv: 2605.00357 · v1 · submitted 2026-05-01 · 💻 cs.GR · cs.HC· cs.MM

Recognition: unknown

Towards Interactive Multimodal Representation of ML Functions for Human Understanding of ML

Bokang Wang , Yingxuan Liao , Leah Lee , Jack Wesson , Anlan Yang , Ruizi Wang , Yigang Wen

Authors on Pith no claims yet

Pith reviewed 2026-05-09 15:27 UTC · model grok-4.3

classification 💻 cs.GR cs.HCcs.MM

keywords interactive visualizationsmachine learning understandingAI attitudescuriosity in MLmultimodal representationspublic engagementML educationtransparent datasets

0 comments

The pith

Interactive visualizations of machine learning can build understanding and reduce fear among teenagers and non-experts.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that interactive visualizations make machine learning accessible by sparking curiosity in people outside the field. Growing reliance on these systems makes it important to move attitudes away from intimidation by the unknown. The authors built three prototypes with transparent datasets to identify what drives engagement and informed views. If the approach works, it could start a cycle of learning that leads to greater public confidence in ML.

Core claim

By employing interactive visualizations of machine learning data with carefully selected highly-transparent datasets, we examine the success factors of engagement required for more informed attitudes on ML less dictated by the fear of the unknown.

What carries the argument

Interactive multimodal representations of ML functions, used to allow visual exploration and test what produces engagement and curiosity.

Load-bearing premise

The interactive visualizations will successfully increase understanding, spark curiosity, and produce attitudinal paradigm-shifts away from fear of the unknown.

What would settle it

A study in which participants who use the visualizations show no greater ML knowledge or more positive attitudes than a control group that does not.

Figures

Figures reproduced from arXiv: 2605.00357 by Anlan Yang, Bokang Wang, Jack Wesson, Leah Lee, Ruizi Wang, Yigang Wen, Yingxuan Liao.

**Figure 1.** Figure 1: K-Means clustering deconstructs user paintings to show clustering in 3D and VR view at source ↗

**Figure 2.** Figure 2: Musical subcomponent analysis used the Fast view at source ↗

**Figure 3.** Figure 3: Comparison of the haptic feeling and the model of view at source ↗

**Figure 5.** Figure 5: Figurative approach connection and curiosity that primes them for deeper engagement with the world of data and machine learning. Through relatable characters, compelling plotlines, and immersive scenarios, we aim to ignite a spark of interest that propels users towards a more profound understanding of the underlying concepts. With five meticulously designed levels, each meticulously calibrated to spotligh… view at source ↗

read the original abstract

Attitudes about artificial intelligence and machine learning are recent victims of endemic misunderstanding; given our increasing reliance on these technologies, the need for widespread understanding and confidence in their use is paramount. To this end, our work seeks to increase understanding in these typically inaccessible topics through interactive visualizations, thereby garnering curiosity in the hopes of kickstarting a cycle of understanding leading to further pursuit of knowledge. We hope this will cyclically shift global attitudes away from the intimidation of the unknown currently plaguing ML. This work explores best practices for supporting curiosity in new technologies, to inspire attitudinal paradigm-shifts. Over three, distinct visualizations of machine learning data, we created prototypes with carefully selected, highly-transparent datasets, to examine the success factors of engagement required for more informed attitudes on ML less dictated by the fear of the unknown. By employing interactive visualizations, we can captivate the interest of teenagers and individuals from diverse fields, encouraging them to explore the fascinating world of machine learning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper describes three interactive visualization prototypes for ML education but offers no evidence that they improve understanding or shift attitudes.

read the letter

This paper is an early design exploration of interactive visualizations to teach machine learning to non-experts, but it stops short of testing whether those visualizations actually work. The authors created three prototypes using transparent datasets to highlight engagement factors like interactivity and clarity. They argue this approach can build curiosity and reduce fear of ML among teenagers and people from other fields. The discussion of best practices for curiosity support is the main content, drawing on design choices to make abstract ML ideas more approachable. What they do well is pick concrete examples and focus on accessibility. Choosing datasets that are easy to understand helps demystify the black box nature of ML models. The prototypes target specific pain points like intimidation from unknown tech, which is a real issue as AI becomes more common. The soft spot is obvious and central: there are no results from any kind of testing. No user studies, no metrics on improved understanding, no before-and-after attitude surveys. The abstract talks about shifting global attitudes and captivating interest, but those outcomes are assumed rather than shown. This makes the paper more of a proposal than a completed piece of work. The stress test note captures it accurately—the claims depend on the unproven idea that these designs will produce the desired cognitive and emotional effects. Readers who might get something out of it are people in human-computer interaction or computer science education who are brainstorming visualization ideas for ML literacy. It could spark some thoughts on what makes a good interactive demo. For a serious referee, though, it probably doesn't qualify yet because the evidential base is missing. The work shows clear thinking about design but doesn't deliver on the promised examination of success factors. I'd recommend holding off on peer review until they add some form of evaluation, even if it's small-scale pilot data. That would turn this into something more substantial.

Referee Report

2 major / 1 minor

Summary. The paper presents three interactive visualization prototypes for machine learning concepts and data, using transparent datasets, with the aim of increasing understanding, sparking curiosity, and shifting attitudes away from fear of ML among teenagers and non-experts. The work describes design choices for engagement and states that the prototypes were created to examine success factors for more informed attitudes.

Significance. If validated through evaluation, the approach could contribute to accessible ML education and public engagement with AI technologies by leveraging interactive graphics. The manuscript currently offers only prototype descriptions without supporting data, so its significance remains prospective rather than demonstrated.

major comments (2)

[Abstract] Abstract: The claims that the visualizations 'can captivate the interest of teenagers' and 'shift global attitudes away from the intimidation of the unknown' rest on unevaluated prototypes. The manuscript provides no user studies, pre/post comprehension measures, curiosity scales, attitude surveys, or any success metrics to support these outcomes.
[Prototype descriptions] Prototype sections: The paper states the visualizations were created 'to examine the success factors of engagement' but reports no results from such examination, no pilot feedback, and no discussion of how the chosen design elements (e.g., interactivity, transparency) were tested for effectiveness in producing understanding or attitudinal change.

minor comments (1)

[Overall structure] The manuscript would benefit from explicit section headings distinguishing design rationale from any intended future evaluation plans.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript describing three interactive visualization prototypes for machine learning concepts. The work focuses on design choices using transparent datasets to support engagement and curiosity among non-experts. We recognize that the current version presents these prototypes without empirical evaluation and will revise the text to better reflect this scope while outlining future assessment plans.

read point-by-point responses

Referee: [Abstract] Abstract: The claims that the visualizations 'can captivate the interest of teenagers' and 'shift global attitudes away from the intimidation of the unknown' rest on unevaluated prototypes. The manuscript provides no user studies, pre/post comprehension measures, curiosity scales, attitude surveys, or any success metrics to support these outcomes.

Authors: We agree that the abstract asserts prospective benefits without supporting data from user studies or metrics. The prototypes were developed as an initial design exploration informed by principles of interactive visualization and accessible education. In revision, we will rewrite the abstract to describe the prototypes, their transparent datasets, and design goals for sparking curiosity, while removing or qualifying the unverified claims about captivating interest and shifting attitudes. We will add a limitations and future work section detailing planned user studies with appropriate measures. revision: yes
Referee: [Prototype descriptions] Prototype sections: The paper states the visualizations were created 'to examine the success factors of engagement' but reports no results from such examination, no pilot feedback, and no discussion of how the chosen design elements (e.g., interactivity, transparency) were tested for effectiveness in producing understanding or attitudinal change.

Authors: The prototype sections explain the rationale for selecting interactive elements and transparent datasets based on existing literature on engagement in learning tools. No formal examination results, pilot feedback, or effectiveness tests are included because this manuscript centers on the prototyping process itself. We will revise these sections to explicitly state the literature basis for the design choices, note the absence of systematic testing in the current work, and indicate that evaluation of impacts on understanding and attitudes is planned as subsequent research. Any informal development observations will be briefly noted for context. revision: partial

Circularity Check

0 steps flagged

No circularity: descriptive design paper with no derivations, predictions, or fitted inputs

full rationale

The manuscript is a design-oriented description of three interactive visualization prototypes using transparent ML datasets. It contains no equations, no parameter fitting, no predictions of quantitative outcomes, and no derivation chain that could reduce to its own inputs. Claims about captivating interest or shifting attitudes are framed as aspirational goals and design motivations rather than results derived from prior steps or self-citations. No load-bearing self-citations, uniqueness theorems, or ansatzes are invoked. The work is self-contained as a prototype exploration and does not rely on any circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the untested premise that interactive visualizations will drive curiosity and attitudinal change. No free parameters, invented entities, or formal axioms are introduced beyond standard HCI assumptions about engagement.

axioms (1)

domain assumption Interactive visualizations using transparent datasets can increase understanding and reduce intimidation of ML
Invoked throughout the abstract as the mechanism for shifting global attitudes.

pith-pipeline@v0.9.0 · 5484 in / 1223 out tokens · 30354 ms · 2026-05-09T15:27:14.300456+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

6 extracted references · 3 canonical work pages

[1]

Umang Bhatt and Alice Xiang and Shubham Sharma and Adrian Weller and Ankur Taly and Yunhan Jia and Joydeep Ghosh and Ruchir Puri and José M. F. Moura and Peter Eckersley. Explainable Machine Learning in Deployment. Conference on Fairness, Accountability, and Transparency. 2020. doi:10.1145/3351095.3375624

work page doi:10.1145/3351095.3375624 2020
[2]

ACM CHI Conference

Identifying the intersections: User experience + research scientist collaboration in a generative machine learning interface , author =. ACM CHI Conference. 2019. doi:10.1145/3290607.3299059

work page doi:10.1145/3290607.3299059 2019
[3]

Embodied Learning in Immersive Smart Spaces

Mirko Gelsomini and Giulia Leonardi and Franca Garzotto. Embodied Learning in Immersive Smart Spaces. ACM CHI Conference. 2020. doi:10.1145/3313831.3376667

work page doi:10.1145/3313831.3376667 2020
[4]

Special issue: Digital Libraries. 1996

1996
[5]

Ultraleap Official Website

Ultraleap.com. Ultraleap Official Website. 2022

2022
[6]

The Art of Game Design

Jesse Schell. The Art of Game Design. 2019

2019