arxiv: 2604.07121 · v1 · submitted 2026-04-08 · 💻 cs.HC · cs.AI

Recognition: unknown

Mixed-Initiative Context: Structuring and Managing Context for Human-AI Collaboration

Haichang Li , Qinshi Zhang , Piaohong Wang , Zhicong Lu

Authors on Pith no claims yet

Pith reviewed 2026-05-10 17:24 UTC · model grok-4.3

classification 💻 cs.HC cs.AI

keywords mixed-initiative contexthuman-AI collaborationcontext managementmulti-turn interactionscontext structuringinteractive objectHCI

0 comments

The pith

Context in human-AI conversations should be treated as an explicit, shared, and editable object that both sides can actively organize rather than a fixed chronological sequence.

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

The paper claims that natural multi-turn exchanges produce contexts with different lifecycles, hierarchies, and relevance levels, yet current systems flatten them into one unchanging log. This flattening leaves abandoned threads, parallel topics, and outdated details in the active window, creating interference that users can only fix indirectly by rephrasing or repeating. The proposed solution reconceptualizes context as Mixed-Initiative Context, an interactive object whose structure, scope, and content both humans and AI can directly inspect, prune, group, or expand. A probe system called Contextify was implemented to let users perform these operations and to observe how people respond when AI also suggests changes. If the approach works, collaboration becomes more controllable and less prone to conflicts that arise from unmanageable history.

Core claim

The paper establishes that reconceptualizing the context formed across multi-turn interactions as an explicit, structured, and manipulable interactive object enables both humans and AI to actively participate in context construction and regulation, replacing the current practice of treating context as a fixed chronological sequence with no mechanism for dynamic organization.

What carries the argument

Mixed-Initiative Context, the reconceptualization of interaction history as an explicit, structured, and manipulable interactive object that both parties can organize and adjust according to task needs.

If this is right

Users gain direct, verifiable ways to remove or isolate specific exchanges instead of relying on indirect prompt edits.
AI systems can propose context adjustments such as grouping related threads or dropping temporary detours.
Parallel topic threads can be maintained separately without polluting the main reasoning window.
Collaboration workflows can change structure mid-task as new priorities emerge without restarting the entire history.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Designers of long-running AI assistants could surface context as a visible, editable panel rather than a hidden token limit.
Evaluation of conversational agents might shift from single-response accuracy to measures of context coherence over many turns.
The same structuring approach could be tested in domains such as collaborative planning or creative writing to see whether explicit control reduces user frustration.

Load-bearing premise

Contexts formed in multi-turn interactions differ enough in lifecycle, hierarchy, and relevance that treating them as one fixed sequence produces interference and conflict that explicit management can resolve.

What would settle it

A side-by-side comparison of the same multi-turn collaboration tasks run once with standard chronological context and once with the mixed-initiative structured version, checking whether users report fewer conflicts from old or parallel topics and complete tasks with less repetition.

Figures

Figures reproduced from arXiv: 2604.07121 by Haichang Li, Piaohong Wang, Qinshi Zhang, Zhicong Lu.

**Figure 1.** Figure 1: Contextify instantiates the Mixed-Initiative Context concept. (1) Conversational System: Top controls navigate or [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

**Figure 2.** Figure 2: The Mixed-Initiative Context framework. Left: Tra [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Figure 3: Flat conversational context collapses heterogeneous task elements into a single linear transcript, making boundaries [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗

read the original abstract

In the human-AI collaboration area, the context formed naturally through multi-turn interactions is typically flattened into a chronological sequence and treated as a fixed whole in subsequent reasoning, with no mechanism for dynamic organization and management along the collaboration workflow. Yet these contexts differ substantially in lifecycle, structural hierarchy, and relevance. For instance, temporary or abandoned exchanges and parallel topic threads persist in the limited context window, causing interference and even conflict. Meanwhile, users are largely limited to influencing context indirectly through input modifications (e.g., corrections, references, or ignoring), leaving their control neither explicit nor verifiable. To address this, we propose Mixed-Initiative Context, which reconceptualizes the context formed across multi-turn interactions as an explicit, structured, and manipulable interactive object. Under this concept, the structure, scope, and content of context can be dynamically organized and adjusted according to task needs, enabling both humans and AI to actively participate in context construction and regulation. To explore this concept, we implement Contextify as a probe system and conduct a user study examining users' context management behaviors, attitudes toward AI initiative, and overall collaboration experience. We conclude by discussing the implications of this concept for the HCI community.

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 frames context in long human-AI chats as an explicit, jointly editable object and tests it with a probe system, but the user study does not isolate whether the structure itself reduces interference.

read the letter

The main contribution is treating multi-turn context not as a fixed chronological list but as a structured object that both the human and the AI can inspect, reorganize, and prune during the session. Contextify implements this with visible threads, scopes, and edit actions, and the user study looks at how participants actually used those controls plus their views on AI initiative. That setup directly targets the problem of abandoned or parallel threads persisting and creating noise later on, which is a real friction in current tools. The paper does a clean job laying out the lifecycle and hierarchy differences that standard logs ignore. The study covers behaviors, attitudes, and experience, which gives a starting picture of how people might respond to the new controls. The soft spot is the missing baseline. Without a condition that uses ordinary chronological context under the same tasks, any reported improvements in perceived conflict or task flow could come from interface novelty, task instructions, or simply having more visible options rather than from the mixed-initiative structuring. The central claim that explicit management resolves interference therefore rests on an assumption that the data do not yet separate from other factors. This work sits squarely in HCI research on AI collaboration tools. Readers who build or study long-horizon interfaces will find the concrete probe and the framing useful to discuss, even if the evaluation needs a tighter control. It is worth sending to peer review so the study design can be strengthened and the idea can be placed against related work on dialogue context and shared representations.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes Mixed-Initiative Context as a reconceptualization of multi-turn human-AI interaction context, treating it as an explicit, structured, and dynamically manipulable object rather than a flattened chronological sequence. It argues that differing lifecycles, hierarchies, and relevance across context elements cause interference and conflict, and that both humans and AI should actively participate in context construction and regulation. The authors implement the idea in the Contextify probe system and conduct a user study to examine context management behaviors, attitudes toward AI initiative, and collaboration experience.

Significance. If the central claims hold, the work could meaningfully advance HCI research on human-AI collaboration by providing a framework for explicit context regulation that reduces interference in long-running interactions. The probe system offers a concrete artifact for exploring mixed-initiative mechanisms, which may inform future designs of controllable and verifiable collaborative interfaces.

major comments (2)

[User Study] User Study section: the evaluation uses only the Contextify probe without a controlled baseline comparison against standard chronological context (unmodified chat interfaces). This leaves open whether observed behaviors, reduced perceived conflict, or positive attitudes stem from the explicit structuring or from novelty, task framing, or demand characteristics; metrics such as task success, edit frequency, or conflict reports against a within- or between-subjects control are needed to isolate the effect.
[Abstract] Abstract and Evaluation: no details are provided on study design (e.g., tasks, participant count, measures, or quantitative results), weakening the empirical support for claims that explicit management resolves interference from differing context lifecycles and hierarchies.

minor comments (2)

The terms 'lifecycle,' 'structural hierarchy,' and 'relevance' of context elements are used without formal definitions or examples; adding a short taxonomy or illustrative scenarios would improve clarity and reproducibility.
[Implementation] The manuscript would benefit from explicit discussion of how Contextify's interface mechanisms (e.g., editing, scoping) map to the proposed concept, including any limitations observed during the probe study.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We address each major comment below, indicating where revisions will be made to strengthen the manuscript while preserving the exploratory nature of the probe study.

read point-by-point responses

Referee: [User Study] User Study section: the evaluation uses only the Contextify probe without a controlled baseline comparison against standard chronological context (unmodified chat interfaces). This leaves open whether observed behaviors, reduced perceived conflict, or positive attitudes stem from the explicit structuring or from novelty, task framing, or demand characteristics; metrics such as task success, edit frequency, or conflict reports against a within- or between-subjects control are needed to isolate the effect.

Authors: We agree that the lack of a controlled baseline comparison is a limitation that prevents strong causal claims about the specific benefits of explicit mixed-initiative context structuring versus confounds such as novelty or demand characteristics. The study was intentionally designed as an exploratory probe investigation to surface user behaviors, attitudes toward AI initiative, and collaboration experiences in this new paradigm, rather than as a comparative experiment. In the revision we will add an explicit limitations subsection that acknowledges this gap, reports available quantitative metrics from the existing data (e.g., edit frequencies and self-reported conflict), and outlines concrete directions for future controlled studies. We will not, however, be able to conduct a new within- or between-subjects baseline experiment at this stage. revision: partial
Referee: [Abstract] Abstract and Evaluation: no details are provided on study design (e.g., tasks, participant count, measures, or quantitative results), weakening the empirical support for claims that explicit management resolves interference from differing context lifecycles and hierarchies.

Authors: We accept that the current abstract provides insufficient detail on the empirical component. In the revised manuscript we will expand the abstract to include the number of participants, the tasks employed, the primary measures (behavioral logs, questionnaires on attitudes and collaboration experience), and key quantitative and qualitative findings. This will give readers a clearer view of the evidence supporting the claims about reduced interference through explicit context management. revision: yes

Circularity Check

0 steps flagged

No circularity: conceptual proposal with independent empirical exploration

full rationale

The paper advances a new conceptual framework (Mixed-Initiative Context) by identifying limitations in existing chronological context handling, then implements Contextify as a probe and reports a user study on behaviors and attitudes. No equations, fitted parameters, predictions, or derivations appear. The central claim is the proposal itself, not a result derived from prior fitted quantities or self-referential theorems. Self-citations, if present, are not load-bearing for the core reconceptualization. The derivation chain is self-contained as a design-oriented contribution rather than a reduction to inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The proposal rests on the domain assumption that current flattened context handling produces interference and that explicit structure will enable better regulation; the new concept itself is introduced without independent empirical validation beyond the probe study.

axioms (1)

domain assumption Contexts in multi-turn human-AI interactions differ substantially in lifecycle, structural hierarchy, and relevance, and flattened chronological treatment causes interference.
Stated directly in the abstract as the motivation for the new concept.

invented entities (1)

Mixed-Initiative Context no independent evidence
purpose: Reconceptualize context as an explicit, structured, manipulable interactive object that humans and AI can dynamically organize.
New framing introduced to address limitations of existing context handling; no independent falsifiable evidence provided in the abstract.

pith-pipeline@v0.9.0 · 5522 in / 1300 out tokens · 35841 ms · 2026-05-10T17:24:40.543079+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

57 extracted references · 42 canonical work pages

[1]

Allen, C.I

J.E. Allen, C.I. Guinn, and E. Horvtz. 1999. Mixed-initiative interaction.IEEE Intelligent Systems and their Applications14, 5 (1999), 14–23. doi:10.1109/5254. 796083

work page doi:10.1109/5254 1999
[2]

Saleema Amershi, Maya Cakmak, William Bradley Knox, and Todd Kulesza. 2014. Power to the People: The Role of Humans in Interactive Machine Learning.AI Magazine35, 4 (Dec. 2014), 105–120. doi:10.1609/aimag.v35i4.2513

work page doi:10.1609/aimag.v35i4.2513 2014
[3]

Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz

Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N. Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz. 2019. Guidelines for Human- AI Interaction. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems(Glasgow, Scotland Uk)(CHI ’19). Associa...

work page doi:10.1145/3290605.3300233 2019
[4]

Tyler Angert, Miroslav Suzara, Jenny Han, Christopher Pondoc, and Hariharan Subramonyam. 2023. Spellburst: A Node-based Interface for Exploratory Creative Coding with Natural Language Prompts. InProceedings of the 36th Annual ACM Symposium on User Interface Software and Technology(San Francisco, CA, USA) (UIST ’23). Association for Computing Machinery, Ne...

work page doi:10.1145/3586183.3606719 2023
[5]

Anysphere. 2026. Cursor: The AI-First Code Editor. https://www.cursor.com/. Accessed: March 31, 2026

2026
[6]

Laurens Boer and Jared Donovan. 2012. Provotypes for participatory innovation. InProceedings of the Designing Interactive Systems Conference(Newcastle Upon Tyne, United Kingdom)(DIS ’12). Association for Computing Machinery, New York, NY, USA, 388–397. doi:10.1145/2317956.2318014

work page doi:10.1145/2317956.2318014 2012
[7]

Yining Cao, Jane L E, Chen Zhu-Tian, and Haijun Xia. 2023. DataParticles: Block- based and Language-oriented Authoring of Animated Unit Visualizations. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (Hamburg, Germany)(CHI ’23). Association for Computing Machinery, New York, NY, USA, Article 808, 15 pages. doi:10.1145/35445...

work page doi:10.1145/3544548.3581472 2023
[8]

John Joon Young Chung, Wooseok Kim, Kang Min Yoo, Hwaran Lee, Eytan Adar, and Minsuk Chang. 2022. TaleBrush: Sketching Stories with Generative Pretrained Language Models. InProceedings of the 2022 CHI Conference on Human Factors in Computing Systems(New Orleans, LA, USA)(CHI ’22). Association for Computing Machinery, New York, NY, USA, Article 209, 19 pag...

work page arXiv 2022
[9]

Herbert H Clark and Susan E Brennan. 1991. Grounding in communication. (1991)

1991
[10]

Adam J Coscia, Shunan Guo, Eunyee Koh, and Alex Endert. 2025. OnGoal: Tracking and Visualizing Conversational Goals in Multi-Turn Dialogue with Large Language Models. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 208, 18 pages. doi:10.114...

work page doi:10.1145/3746059 2025
[11]

2005.A Study of the Design Process - The Double Diamond

Design Council. 2005.A Study of the Design Process - The Double Diamond. Technical Report. Design Council. https://www.designcouncil.org.uk/our- resources/the-double-diamond/ Accessed: 2026-03-31

2005
[12]

Paul Dourish. 2004. What we talk about when we talk about context.Personal Ubiquitous Comput.8, 1 (Feb. 2004), 19–30. doi:10.1007/s00779-003-0253-8

work page doi:10.1007/s00779-003-0253-8 2004
[13]

Jerry Alan Fails and Dan R. Olsen. 2003. Interactive machine learning. InPro- ceedings of the 8th International Conference on Intelligent User Interfaces(Miami, Florida, USA)(IUI ’03). Association for Computing Machinery, New York, NY, USA, 39–45. doi:10.1145/604045.604056

work page doi:10.1145/604045.604056 2003
[14]

Fussell, Robert E

Susan R. Fussell, Robert E. Kraut, and Jane Siegel. 2000. Coordination of commu- nication: effects of shared visual context on collaborative work. InProceedings of the 2000 ACM Conference on Computer Supported Cooperative Work(Philadelphia, Pennsylvania, USA)(CSCW ’00). Association for Computing Machinery, New York, NY, USA, 21–30. doi:10.1145/358916.358947

work page doi:10.1145/358916.358947 2000
[15]

Bill Gaver, Tony Dunne, and Elena Pacenti. 1999. Design: Cultural probes. Interactions6, 1 (Jan. 1999), 21–29. doi:10.1145/291224.291235

work page doi:10.1145/291224.291235 1999
[16]

Chakraborti, C

Katy Ilonka Gero, Chelse Swoopes, Ziwei Gu, Jonathan K. Kummerfeld, and Elena L. Glassman. 2024. Supporting Sensemaking of Large Language Model Outputs at Scale. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 838, 21 pages. doi:10.1...

work page doi:10.1145/3613904 2024
[17]

Saul Greenberg. 2001. Context as a dynamic construct.Hum.-Comput. Interact. 16, 2 (Dec. 2001), 257–268. doi:10.1207/S15327051HCI16234_09

work page doi:10.1207/s15327051hci16234_09 2001
[18]

Saul Greenberg and Bill Buxton. 2008. Usability evaluation considered harmful (some of the time). InProceedings of the SIGCHI Conference on Human Factors in Computing Systems(Florence, Italy)(CHI ’08). Association for Computing Machinery, New York, NY, USA, 111–120. doi:10.1145/1357054.1357074

work page doi:10.1145/1357054.1357074 2008
[19]

Ken Gu, Ruoxi Shang, Tim Althoff, Chenglong Wang, and Steven M. Drucker
[20]

I want it to talk like Darth Vader

How Do Analysts Understand and Verify AI-Assisted Data Analyses?. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 748, 22 pages. doi:10.1145/3613904.3642497

work page doi:10.1145/3613904.3642497 2024
[21]

Ziyao He, Yunpeng Song, Shurui Zhou, and Zhongmin Cai. 2023. Interaction of Thoughts: Towards Mediating Task Assignment in Human-AI Cooperation with a Capability-Aware Shared Mental Model. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems(Hamburg, Germany)(CHI ’23). Association for Computing Machinery, New York, NY, USA, Artic...

work page doi:10.1145/3544548.3580983 2023
[22]

Marti A Hearst, J Allen, C Guinn, and Eric Horvitz. 1999. Mixed-initiative interaction: Trends and controversies.IEEE Intelligent Systems14, 5 (1999), 14–23

1999
[23]

James Hollan, Edwin Hutchins, and David Kirsh. 2000. Distributed cognition: toward a new foundation for human-computer interaction research.ACM Trans. Comput.-Hum. Interact.7, 2 (June 2000), 174–196. doi:10.1145/353485.353487

work page doi:10.1145/353485.353487 2000
[24]

Shelton, Fanny Chevalier, Kari Kraus, and Niklas Elmqvist

Md Naimul Hoque, Tasfia Mashiat, Bhavya Ghai, Cecilia D. Shelton, Fanny Chevalier, Kari Kraus, and Niklas Elmqvist. 2024. The HaLLMark Effect: Sup- porting Provenance and Transparent Use of Large Language Models in Writing with Interactive Visualization. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ...

work page doi:10.1145/3613904.3641895 2024
[25]

Eric Horvitz. 1999. Principles of mixed-initiative user interfaces. InProceedings of the SIGCHI Conference on Human Factors in Computing Systems(Pittsburgh, Pennsylvania, USA)(CHI ’99). Association for Computing Machinery, New York, NY, USA, 159–166. doi:10.1145/302979.303030

work page doi:10.1145/302979.303030 1999
[26]

Bederson, Al- lison Druin, Catherine Plaisant, Michel Beaudouin-Lafon, Stéphane Conversy, Helen Evans, Heiko Hansen, Nicolas Roussel, and Björn Eiderbäck

Hilary Hutchinson, Wendy Mackay, Bo Westerlund, Benjamin B. Bederson, Al- lison Druin, Catherine Plaisant, Michel Beaudouin-Lafon, Stéphane Conversy, Helen Evans, Heiko Hansen, Nicolas Roussel, and Björn Eiderbäck. 2003. Technol- ogy probes: inspiring design for and with families. InProceedings of the SIGCHI Conference on Human Factors in Computing System...

work page doi:10.1145/642611.642616 2003
[27]

Dow, and Haijun Xia

Peiling Jiang, Jude Rayan, Steven P. Dow, and Haijun Xia. 2023. Graphologue: Exploring Large Language Model Responses with Interactive Diagrams. InPro- ceedings of the 36th Annual ACM Symposium on User Interface Software and Technology(San Francisco, CA, USA)(UIST ’23). Association for Computing Ma- chinery, New York, NY, USA, Article 3, 20 pages. doi:10....

work page doi:10.1145/3586183.3606737 2023
[28]

Team Player

Gary Klein, David D. Woods, Jeffrey M. Bradshaw, Robert R. Hoffman, and Paul J. Feltovich. 2004. Ten Challenges for Making Automation a "Team Player" in Joint Human-Agent Activity.IEEE Intelligent Systems19, 6 (Nov. 2004), 91–95. doi:10.1109/MIS.2004.74

work page doi:10.1109/mis.2004.74 2004
[29]

Mina Lee, Percy Liang, and Qian Yang. 2022. CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems (New Orleans, LA, USA)(CHI ’22). Association for Computing Machinery, New York, NY, USA, Article 388, 19 pages. doi:10.1145/34911...

work page doi:10.1145/3491102.3502030 2022
[30]

Florian Lehmann. 2023. Mixed-Initiative Interaction with Computational Gen- erative Systems. InExtended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems(Hamburg, Germany)(CHI EA ’23). Associa- tion for Computing Machinery, New York, NY, USA, Article 501, 6 pages. doi:10.1145/3544549.3577061

work page doi:10.1145/3544549.3577061 2023
[31]

Haichang Li, Anjun Zhu, and Arpit Narechania. 2026. Alignment- Process-Outcome: Rethinking How AIs and Humans Collaborate. arXiv:2603.08017 [cs.HC] https://arxiv.org/abs/2603.08017

work page arXiv 2026
[32]

Vera and Vaughan, Jennifer Wortman , year =

Q. Vera Liao and Jennifer Wortman Vaughan. 2023. AI Transparency in the Age of LLMs: A Human-Centered Research Roadmap. arXiv:2306.01941 [cs.HC] https://arxiv.org/abs/2306.01941

work page arXiv 2023
[33]

Youn-Kyung Lim, Erik Stolterman, and Josh Tenenberg. 2008. The anatomy of prototypes: Prototypes as filters, prototypes as manifestations of design ideas. ACM Trans. Comput.-Hum. Interact.15, 2, Article 7 (July 2008), 27 pages. doi:10. 1145/1375761.1375762

work page arXiv 2008
[34]

Atefeh Mahdavi Goloujeh, Anne Sullivan, and Brian Magerko. 2024. Is It AI or Is It Me? Understanding Users’ Prompt Journey with Text-to-Image Generative AI Tools. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 183, 13 pages. doi:10....

work page doi:10.1145/3613904.3642861 2024
[35]

Damien Masson, Sylvain Malacria, Géry Casiez, and Daniel Vogel. 2024. Direct- GPT: A Direct Manipulation Interface to Interact with Large Language Models. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems Li et al. (Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 975, 16 pages. doi:...

work page doi:10.1145/3613904.3642462 2024
[36]

Robert Nimmo, Marios Constantinides, Ke Zhou, Daniele Quercia, and Simone Stumpf. 2024. User Characteristics in Explainable AI: The Rabbit Hole of Personal- ization?. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 317, 13 pages. doi...

work page doi:10.1145/3613904.3642352 2024
[37]

Helen Nissenbaum. 2004. Privacy as contextual integrity.Wash. L. Rev.79 (2004), 119

2004
[38]

D Norman. 1988. Design Of Everyday Things

1988
[39]

OpenAI. 2023. ChatGPT. https://chat.openai.com/chat. Accessed: March 31, 2026

2023
[40]

James Pierce and Eric Paulos. 2015. Making Multiple Uses of the Obscura 1C Digital Camera: Reflecting on the Design, Production, Packaging and Distribution of a Counterfunctional Device. InProceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems(Seoul, Republic of Korea)(CHI ’15). Association for Computing Machinery, New York, N...

work page arXiv 2015
[41]

Peter Pirolli and Stuart Card. 2005. The sensemaking process and leverage points for analyst technology as identified through cognitive task analysis. In Proceedings of international conference on intelligence analysis, Vol. 5. McLean, VA, USA, 2–4

2005
[42]

Dimitri Popolov, Michael Callaghan, and Paul Luker. 2000. Conversation space: visualising multi-threaded conversation. InProceedings of the Working Confer- ence on Advanced Visual Interfaces(Palermo, Italy)(A VI ’00). Association for Computing Machinery, New York, NY, USA, 246–249. doi:10.1145/345513.345330

work page doi:10.1145/345513.345330 2000
[43]

Donald A. Schön. 1983.The Reflective Practitioner: How Professionals Think in Action. Basic Books, New York

1983
[44]

Bernstein

Omar Shaikh, Shardul Sapkota, Shan Rizvi, Eric Horvitz, Joon Sung Park, Diyi Yang, and Michael S. Bernstein. 2025. Creating General User Models from Com- puter Use. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 35, 23 pages. doi:10.1145/3...

work page doi:10.1145/3746059.3747722 2025
[45]

Ben Shneiderman and Pattie Maes. 1997. Direct manipulation vs. interface agents. Interactions4, 6 (Nov. 1997), 42–61. doi:10.1145/267505.267514

work page doi:10.1145/267505.267514 1997
[46]

Hari Subramonyam, Roy Pea, Christopher Pondoc, Maneesh Agrawala, and Colleen Seifert. 2024. Bridging the Gulf of Envisioning: Cognitive Challenges in Prompt Based Interactions with LLMs. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Asso- ciation for Computing Machinery, New York, NY, USA, Arti...

work page doi:10.1145/3613904.3642754 2024
[47]

1987.Plans and situated actions: The problem of human- machine communication

Lucille Alice Suchman. 1987.Plans and situated actions: The problem of human- machine communication. Cambridge university press

1987
[48]

Sangho Suh, Bryan Min, Srishti Palani, and Haijun Xia. 2023. Sensecape: En- abling Multilevel Exploration and Sensemaking with Large Language Models. InProceedings of the 36th Annual ACM Symposium on User Interface Software and Technology(San Francisco, CA, USA)(UIST ’23). Association for Computing Machinery, New York, NY, USA, Article 1, 18 pages. doi:10...

work page doi:10.1145/3586183.3606756 2023
[49]

Michelle Vaccaro, Abdullah Almaatouq, and Thomas Malone. 2024. When com- binations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour8, 12 (2024), 2293–2303

2024
[50]

It Felt Like Having a Second Mind

Qian Wan, Siying Hu, Yu Zhang, Piaohong Wang, Bo Wen, and Zhicong Lu. 2024. "It Felt Like Having a Second Mind": Investigating Human-AI Co-creativity in Prewriting with Large Language Models.Proc. ACM Hum.-Comput. Interact.8, CSCW1, Article 84 (April 2024), 26 pages. doi:10.1145/3637361

work page doi:10.1145/3637361 2024
[51]

Qiaosi Wang and Ashok K. Goel. 2024. Mutual Theory of Mind for Human-AI Communication. arXiv:2210.03842 [cs.HC] https://arxiv.org/abs/2210.03842

work page arXiv 2024
[52]

Liwenhan Xie, Chengbo Zheng, Haijun Xia, Huamin Qu, and Chen Zhu-Tian
[53]

InProceedings of the 37th Annual ACM Symposium on User Interface Software and Technology(Pittsburgh, PA, USA) (UIST ’24)

WaitGPT: Monitoring and Steering Conversational LLM Agent in Data Analysis with On-the-Fly Code Visualization. InProceedings of the 37th Annual ACM Symposium on User Interface Software and Technology(Pittsburgh, PA, USA) (UIST ’24). Association for Computing Machinery, New York, NY, USA, Article 119, 14 pages. doi:10.1145/3654777.3676374

work page doi:10.1145/3654777.3676374
[54]

for his investigations of the densities of the most important gases and for his discovery of argon in connection with these studies

Hamed Zamani, Johanne R. Trippas, Jeff Dalton, and Filip Radlinski. 2023. Con- versational Information Seeking. arXiv:2201.08808 [cs.IR] https://arxiv.org/abs/ 2201.08808

work page arXiv 2023
[55]

When to Hand Off, When to Work Together

Jiayi Zhou, Renzhong Li, Junxiu Tang, Tan Tang, Haotian Li, Weiwei Cui, and Yingcai Wu. 2024. Understanding Nonlinear Collaboration between Human and AI Agents: A Co-design Framework for Creative Design. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI ’24). Association for Computing Machinery, New York...

work page doi:10.1145/3613904.3642812 2024
[56]

A primary structural action: - continue - branch - return_parent
[57]

easy to answer inline

An optional asset action: - none - extract_reasoning - extract_task_sop Your default action is continue. (Structure agent system prompt, continued.) General principles: - Minimize interruption. - If there is meaningful uncertainty, choose continue. - A missed suggestion is often better than an annoying or premature suggestion. - Optimize for the user's pr...