Contextualized Dynamic Explanations: A Vision

Greg Briskin; Jason H Li; Zhicheng Liu

arxiv: 2605.18698 · v1 · pith:GENG3UV7new · submitted 2026-05-18 · 💻 cs.HC

Contextualized Dynamic Explanations: A Vision

Zhicheng Liu , Jason H Li , Greg Briskin This is my paper

Pith reviewed 2026-05-20 08:25 UTC · model grok-4.3

classification 💻 cs.HC

keywords audienceexplanationscontextualizedcodexcommunicationdata-drivendynamiceffective

0 comments

The pith

Autonomous agents can generate dynamic explanations that adapt to the audience's evolving understanding and communication context.

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

The paper proposes Contextualized Dynamic Explanations (CODEX) as a way to address the shortcomings of static, asynchronous data explanations. These traditional methods often fail to tailor content to the specific audience or allow for active engagement because creators cannot foresee all possible interaction scenarios. Instead, CODEX relies on autonomous agents that build and update a model of the audience while pursuing a set communication goal to create fitting multi-modal presentations. This vision matters because it could transform how complex data is communicated in real-time interactive environments, leading to better comprehension and engagement.

Core claim

The central claim is that it is impossible for communicators to anticipate the full range of interactive scenarios involving the target audience. This motivates the development of autonomous agents capable of evaluating communication progress, making context-sensitive decisions, and producing effective information representations using an evolving audience model and a predefined communication intent.

What carries the argument

The Contextualized Dynamic Explanations (CODEX) framework, which is an agentic approach to dynamically generating contextualized multi-modal information interfaces based on an evolving audience model and predefined communication intent.

Load-bearing premise

Autonomous agents are able to accurately track and model the audience's understanding while making reliable decisions that enhance explanation quality without any human supervision.

What would settle it

A controlled user study comparing audience comprehension and engagement levels when using static explanations versus those generated by an autonomous CODEX agent, with no significant difference found in favor of the dynamic approach.

Figures

Figures reproduced from arXiv: 2605.18698 by Greg Briskin, Jason H Li, Zhicheng Liu.

read the original abstract

Asynchronous data-driven explanations often fail because the content and presentation are not tailored to the target audience, and they provide limited opportunities for active audience engagement. We present a vision for Contextualized Dynamic Explanations (CODEX), an agentic approach to dynamically generating contextualized multi-modal information interfaces for effective data-driven explanations based on an evolving audience model and a predefined communication intent. The premise underlying CODEX is that it is impossible for communicators to anticipate the full range of interactive scenarios involving the target audience. This observation motivates a set of research challenges focused on developing autonomous agents capable of evaluating communication progress, making context-sensitive decisions, and producing effective information representations.

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

This is a vision paper sketching an agent-based framework for adaptive explanations but without any prototypes or tests to back it up.

read the letter

The main point is that this paper lays out a vision called CODEX for using autonomous agents to generate dynamic, multi-modal explanations that adapt based on an evolving model of the audience and a fixed communication goal. The authors argue that static or pre-planned explanations cannot cover the full range of possible interactions, which leads them to list research challenges around tracking progress, making context-aware choices, and producing better representations on the fly. That framing pulls together threads from adaptive interfaces and agentic AI in a straightforward way that could help focus future work in HCI and explainable AI. It does a reasonable job stating the motivation clearly and breaking the problem into actionable areas without overclaiming results. The absence of any implementation, data, or even a small worked example is the obvious limitation. The proposal rests on the idea that agents can reliably infer and update audience understanding in real time, yet the paper offers no discussion of how that modeling would be done or validated. As a result the challenges feel broad and the feasibility remains open. This is the kind of piece that would interest researchers already working on personalized or interactive explanations who want to see a high-level research agenda. Readers seeking methods, algorithms, or empirical findings will not find them here. The thinking is coherent on its own terms and the challenges are stated plainly enough that it merits peer review so the community can comment on priorities and next steps.

Referee Report

0 major / 1 minor

Summary. The manuscript outlines a vision for Contextualized Dynamic Explanations (CODEX), an agent-based system designed to create dynamic, context-sensitive, multi-modal explanations for data-driven content. The approach relies on an evolving model of the audience's understanding and a fixed communication goal. The authors argue that the impossibility of foreseeing all possible audience interactions justifies the use of autonomous agents to assess explanation effectiveness and adapt the presentation accordingly.

Significance. This vision paper identifies a promising direction for improving explanatory interfaces in human-computer interaction. By focusing on agentic capabilities for audience modeling and adaptive decision-making, it highlights opportunities to overcome the limitations of static explanations. The enumeration of research challenges provides a useful framework for future studies, even in the absence of implemented prototypes or empirical results.

minor comments (1)

[Abstract] Consider expanding on how the 'predefined communication intent' is specified and maintained throughout the interaction to clarify the scope of agent autonomy.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of our vision paper and for recommending minor revision. The summary accurately captures the core premise of CODEX: that the impossibility of anticipating all audience interactions motivates an agentic approach to dynamic, context-sensitive explanations based on an evolving audience model and fixed communication goal. We are pleased that the work is viewed as identifying a promising direction and providing a useful framework for future studies.

Circularity Check

0 steps flagged

No significant circularity in conceptual vision paper

full rationale

The paper is a vision piece that states a motivating premise (impossibility of anticipating all interactive scenarios) and enumerates open research challenges for autonomous agents in audience modeling and context-sensitive explanation generation. No equations, fitted parameters, derivations, or self-citations appear in the provided text. The central argument does not reduce to any input by construction and remains a self-contained call for future work rather than a technical result whose validity depends on internal circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The work rests on a single domain assumption about the unpredictability of audience interactions and introduces one new conceptual entity without independent evidence.

axioms (1)

domain assumption It is impossible for communicators to anticipate the full range of interactive scenarios involving the target audience.
This premise is stated directly in the abstract as the motivation for the entire CODEX approach.

invented entities (1)

CODEX autonomous agents no independent evidence
purpose: To evaluate communication progress, make context-sensitive decisions, and produce effective multi-modal information representations.
New agent concept introduced to operationalize the vision; no independent evidence or falsifiable prediction is supplied.

pith-pipeline@v0.9.0 · 5631 in / 1208 out tokens · 46683 ms · 2026-05-20T08:25:08.762377+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The premise underlying CODEX is that it is impossible for communicators to anticipate the full range of interactive scenarios involving the target audience. This observation motivates a set of research challenges focused on developing autonomous agents capable of evaluating communication progress, making context-sensitive decisions, and producing effective information representations.
IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean absolute_floor_iff_bare_distinguishability unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Heterogeneous models... audience model describing the traits, knowledge, and preferences of the target audience.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages · 1 internal anchor

[1]

E. Adar, C. Gearig, A. Balasubramanian, and J. Hullman. Persa- Log: Personalization of News Article Content. InProceedings of the 2017 CHI Conference on Human Factors in Computing Systems, pp. 3188–3200. ACM, Denver Colorado USA, May 2017. doi: 10.1145/ 3025453.3025631 4

work page arXiv 2017
[2]

Adar and E

E. Adar and E. Lee. Communicative visualizations as a learning prob- lem.IEEE Transactions on Visualization and Computer Graphics, 27(2):946–956, 2020. Publisher: IEEE. 3

work page 2020
[3]

Amini, N

F. Amini, N. Henry Riche, B. Lee, C. Hurter, and P. Irani. Understand- ing Data Videos: Looking at Narrative Visualization through the Cin- ematography Lens. InProceedings of the 33rd Annual ACM Confer- ence on Human Factors in Computing Systems, pp. 1459–1468. ACM, Apr. 2015. doi: 10.1145/2702123.2702431 4

work page doi:10.1145/2702123.2702431 2015
[4]

Explorable Explanations

Bret Victor. Explorable Explanations. https://worrydream.com/ExplorableExplanations/. 1, 4

work page
[5]

Brusilovsky and E

P. Brusilovsky and E. Mill ´an. User Models for Adaptive Hypermedia and Adaptive Educational Systems. In P. Brusilovsky, A. Kobsa, and W. Nejdl, eds.,The Adaptive Web: Methods and Strategies of Web Personalization, pp. 3–53. Springer, Berlin, Heidelberg, 2007. doi: 10 .1007/978-3-540-72079-9 1 3

work page 2007
[6]

Collaris and J

D. Collaris and J. J. Van Wijk. ExplainExplore: Visual exploration of machine learning explanations. In2020 IEEE Pacific Visualization Symposium (PacificVis), pp. 26–35. IEEE, 2020. 1

work page 2020
[7]

Dragicevic, Y

P. Dragicevic, Y . Jansen, A. Sarma, M. Kay, and F. Chevalier. In- creasing the transparency of research papers with explorable multi- verse analyses. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems, pp. 1–15, 2019. 1

work page 2019
[8]

S. Geng, Z. Fu, Y . Ge, L. Li, G. De Melo, and Y . Zhang. Improving personalized explanation generation through visualization. InPro- ceedings of the 60th Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pp. 244–255, 2022. 4

work page 2022
[9]

Hohman, M

F. Hohman, M. Conlen, J. Heer, and D. H. Chau. Communicating with interactive articles.Distill, 2020. doi: 10.23915/distill.00028 4

work page doi:10.23915/distill.00028 2020
[10]

F. C. Keil. Explanation and Understanding.Annual Review of Psy- chology, 57(1):227–254, Jan. 2006. doi: 10.1146/annurev.psych.57. 102904.190100 1

work page doi:10.1146/annurev.psych.57 2006
[11]

Mohseni, N

S. Mohseni, N. Zarei, and E. D. Ragan. A Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems. ACM Transactions on Interactive Intelligent Systems, 11(3-4):1–45, Dec. 2021. doi: 10.1145/3387166 4

work page doi:10.1145/3387166 2021
[12]

Memory — OpenAI Help Center

OpenAI. Memory — OpenAI Help Center. https://help.openai.com/en/collections/8471548-memory. 3

work page arXiv
[13]

What Is Cyber Threat Intelligence (CTI)? https://www.paloaltonetworks.com/cyberpedia/what-is-cyberthreat- intelligence-cti

Palto Alto Networks. What Is Cyber Threat Intelligence (CTI)? https://www.paloaltonetworks.com/cyberpedia/what-is-cyberthreat- intelligence-cti. 2

work page
[14]

Personalized explanation in machine learning: A conceptualization

J. Schneider and J. Handali. Personalized explanation in machine learning: A conceptualization, Apr. 2019. arXiv:1901.00770 [cs]. doi: 10.48550/arXiv.1901.00770 4

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.00770 2019
[15]

Segel and J

E. Segel and J. Heer. Narrative visualization: Telling stories with data.IEEE transactions on visualization and computer graphics, 16(6):1139–1148, 2010. Publisher: IEEE. 4

work page 2010
[16]

Shaikh, S

O. Shaikh, S. Sapkota, S. Rizvi, E. Horvitz, J. S. Park, D. Yang, and M. S. Bernstein. Creating general user models from computer use. arXiv preprint arXiv:2505.10831, 2025. 2, 3

work page arXiv 2025
[17]

S. Shin, I. Na, and N. Elmqvist. Drillboards: Adaptive Visualiza- tion Dashboards for Dynamic Personalization of Visualization Expe- riences.IEEE Transactions on Visualization and Computer Graphics,

work page
[18]

Toker, C

D. Toker, C. Conati, G. Carenini, and M. Haraty. Towards Adaptive Information Visualization: On the Influence of User Characteristics. In D. Hutchison, T. Kanade, J. Kittler, J. M. Kleinberg, F. Mattern, J. C. Mitchell, M. Naor, O. Nierstrasz, C. Pandu Rangan, B. Steffen, M. Sudan, D. Terzopoulos, D. Tygar, M. Y . Vardi, G. Weikum, J. Mas- thoff, B. Mobas...

work page doi:10.1007/978-3-642-31454 2012

[1] [1]

E. Adar, C. Gearig, A. Balasubramanian, and J. Hullman. Persa- Log: Personalization of News Article Content. InProceedings of the 2017 CHI Conference on Human Factors in Computing Systems, pp. 3188–3200. ACM, Denver Colorado USA, May 2017. doi: 10.1145/ 3025453.3025631 4

work page arXiv 2017

[2] [2]

Adar and E

E. Adar and E. Lee. Communicative visualizations as a learning prob- lem.IEEE Transactions on Visualization and Computer Graphics, 27(2):946–956, 2020. Publisher: IEEE. 3

work page 2020

[3] [3]

Amini, N

F. Amini, N. Henry Riche, B. Lee, C. Hurter, and P. Irani. Understand- ing Data Videos: Looking at Narrative Visualization through the Cin- ematography Lens. InProceedings of the 33rd Annual ACM Confer- ence on Human Factors in Computing Systems, pp. 1459–1468. ACM, Apr. 2015. doi: 10.1145/2702123.2702431 4

work page doi:10.1145/2702123.2702431 2015

[4] [4]

Explorable Explanations

Bret Victor. Explorable Explanations. https://worrydream.com/ExplorableExplanations/. 1, 4

work page

[5] [5]

Brusilovsky and E

P. Brusilovsky and E. Mill ´an. User Models for Adaptive Hypermedia and Adaptive Educational Systems. In P. Brusilovsky, A. Kobsa, and W. Nejdl, eds.,The Adaptive Web: Methods and Strategies of Web Personalization, pp. 3–53. Springer, Berlin, Heidelberg, 2007. doi: 10 .1007/978-3-540-72079-9 1 3

work page 2007

[6] [6]

Collaris and J

D. Collaris and J. J. Van Wijk. ExplainExplore: Visual exploration of machine learning explanations. In2020 IEEE Pacific Visualization Symposium (PacificVis), pp. 26–35. IEEE, 2020. 1

work page 2020

[7] [7]

Dragicevic, Y

P. Dragicevic, Y . Jansen, A. Sarma, M. Kay, and F. Chevalier. In- creasing the transparency of research papers with explorable multi- verse analyses. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems, pp. 1–15, 2019. 1

work page 2019

[8] [8]

S. Geng, Z. Fu, Y . Ge, L. Li, G. De Melo, and Y . Zhang. Improving personalized explanation generation through visualization. InPro- ceedings of the 60th Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pp. 244–255, 2022. 4

work page 2022

[9] [9]

Hohman, M

F. Hohman, M. Conlen, J. Heer, and D. H. Chau. Communicating with interactive articles.Distill, 2020. doi: 10.23915/distill.00028 4

work page doi:10.23915/distill.00028 2020

[10] [10]

F. C. Keil. Explanation and Understanding.Annual Review of Psy- chology, 57(1):227–254, Jan. 2006. doi: 10.1146/annurev.psych.57. 102904.190100 1

work page doi:10.1146/annurev.psych.57 2006

[11] [11]

Mohseni, N

S. Mohseni, N. Zarei, and E. D. Ragan. A Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems. ACM Transactions on Interactive Intelligent Systems, 11(3-4):1–45, Dec. 2021. doi: 10.1145/3387166 4

work page doi:10.1145/3387166 2021

[12] [12]

Memory — OpenAI Help Center

OpenAI. Memory — OpenAI Help Center. https://help.openai.com/en/collections/8471548-memory. 3

work page arXiv

[13] [13]

What Is Cyber Threat Intelligence (CTI)? https://www.paloaltonetworks.com/cyberpedia/what-is-cyberthreat- intelligence-cti

Palto Alto Networks. What Is Cyber Threat Intelligence (CTI)? https://www.paloaltonetworks.com/cyberpedia/what-is-cyberthreat- intelligence-cti. 2

work page

[14] [14]

Personalized explanation in machine learning: A conceptualization

J. Schneider and J. Handali. Personalized explanation in machine learning: A conceptualization, Apr. 2019. arXiv:1901.00770 [cs]. doi: 10.48550/arXiv.1901.00770 4

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.00770 2019

[15] [15]

Segel and J

E. Segel and J. Heer. Narrative visualization: Telling stories with data.IEEE transactions on visualization and computer graphics, 16(6):1139–1148, 2010. Publisher: IEEE. 4

work page 2010

[16] [16]

Shaikh, S

O. Shaikh, S. Sapkota, S. Rizvi, E. Horvitz, J. S. Park, D. Yang, and M. S. Bernstein. Creating general user models from computer use. arXiv preprint arXiv:2505.10831, 2025. 2, 3

work page arXiv 2025

[17] [17]

S. Shin, I. Na, and N. Elmqvist. Drillboards: Adaptive Visualiza- tion Dashboards for Dynamic Personalization of Visualization Expe- riences.IEEE Transactions on Visualization and Computer Graphics,

work page

[18] [18]

Toker, C

D. Toker, C. Conati, G. Carenini, and M. Haraty. Towards Adaptive Information Visualization: On the Influence of User Characteristics. In D. Hutchison, T. Kanade, J. Kittler, J. M. Kleinberg, F. Mattern, J. C. Mitchell, M. Naor, O. Nierstrasz, C. Pandu Rangan, B. Steffen, M. Sudan, D. Terzopoulos, D. Tygar, M. Y . Vardi, G. Weikum, J. Mas- thoff, B. Mobas...

work page doi:10.1007/978-3-642-31454 2012