pith. sign in

arxiv: 2605.23193 · v1 · pith:XHRMXSREnew · submitted 2026-05-22 · 💻 cs.HC · cs.CL· cs.CY· cs.MA

CultivAgents: Cultivating Relationship-Centered Multi-Agent Systems for Personalized Gardening

Yiyang Wang , Moeiini Reilly , Britney Johnson , Kefei Yan , Alex Cabral , Josiah Hester This is my paper

Pith reviewed 2026-05-25 03:55 UTC · model grok-4.3

classification 💻 cs.HC cs.CLcs.CYcs.MA
keywords multi-agent systemspersonalized gardeningethics of carecommunity gardeningcultural knowledgefood sovereigntyAI for well-being
0
0 comments X

The pith

CultivAgents coordinates three specialized agents to give gardeners skill-adapted, locally grounded, and culturally connected advice that raises their confidence and trust.

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

The paper introduces CultivAgents to address the limits of generic digital gardening tools that overlook users' skills, local ecologies, seasons, and cultural contexts. It proposes a relationship-centered multi-agent system grounded in ethics of care, where an Experience Agent adapts to skill levels, an Environmental Agent incorporates local and seasonal conditions, and an Ethnobotanical Agent links plants to cultural knowledge. A mixed-methods evaluation with community gardeners found reported increases in confidence, motivation, and trust in acting on the advice, indicating the system supports turning interest into situated action for well-being and food autonomy.

Core claim

CultivAgents is a multi-agent system that coordinates an Experience Agent adapting guidance to users' skill levels, an Environmental Agent grounding advice in local and seasonal conditions, and an Ethnobotanical Agent connecting plants to cultural knowledge and histories; evaluation through expert feedback, pre/post surveys, and participatory design shows community gardeners reporting increased confidence from 3.00 to 3.60, motivation from 4.00 to 4.40, and trust in acting on AI advice from 3.20 to 4.00.

What carries the argument

Coordination of three specialized agents—an Experience Agent, an Environmental Agent, and an Ethnobotanical Agent—operating under an ethics of care framework to deliver personalized, socio-culturally grounded gardening support.

If this is right

  • Gardeners gain higher trust when advice accounts for their individual skills and local ecology rather than offering generic recommendations.
  • Complementary perspectives from multiple agents help users translate interest into concrete gardening actions.
  • Multi-agent systems grounded in cultural knowledge can support food sovereignty and community resilience better than single-agent tools.
  • Design of AI for personal activities benefits from explicit coordination between skill adaptation, environmental data, and ethnobotanical context.

Where Pith is reading between the lines

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

  • Similar agent coordination could be applied to other domains involving personal skills and local knowledge, such as meal planning or home maintenance.
  • Addressing the noted limits in cultural specificity would require expanding the ethnobotanical agent's data sources to more diverse traditions.
  • Longer-term deployment studies could test whether the reported gains in motivation lead to sustained changes in gardening behavior.
  • The approach highlights a path for AI systems to preserve cultural practices by embedding them directly into decision support.

Load-bearing premise

Changes in self-reported confidence, motivation, and trust among five community gardeners can be attributed to the CultivAgents system rather than other factors.

What would settle it

A follow-up study with a control group using generic gardening advice that shows no difference in pre/post confidence or trust score changes compared to the CultivAgents group.

Figures

Figures reproduced from arXiv: 2605.23193 by Alex Cabral, Britney Johnson, Josiah Hester, Kefei Yan, Moeiini Reilly, Yiyang Wang.

Figure 1
Figure 1. Figure 1: CultivAgents coordinates experience, environmental, and ethnobotanical agents to provide personalized gardening [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Onboarding dialog collecting gardening experience [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Chat with color-coded agent messages: green (Ex [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Participant-designed CultivAgents personas along an advisor–companion spectrum, including companion-like [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Gardening is critical to support well-being, cultural continuity, and food autonomy, yet existing digital tools often provide generic advice that overlooks gardeners' skills, local ecologies, seasons, and cultural contexts. We introduce CultivAgents, a relationship-centered multi-agent system for personalized, socio-culturally grounded gardening support. Grounded in ethics of care, CultivAgents coordinates multiple specialized agents: an Experience Agent that adapts guidance to users' skill levels, an Environmental Agent that grounds advice in local and seasonal conditions, and an Ethnobotanical Agent that connects plants to cultural knowledge and histories. We evaluated CultivAgents through a three-phase mixed-methods study with domain experts (n=3), HCI researchers (n=7), and community gardeners (n=5), analyzing expert feedback, pre/post surveys, and participatory design activities. Results suggest that CultivAgents helped gardeners translate interest into situated action: community gardeners reported increased confidence (3.00 to 3.60), motivation (4.00 to 4.40), and trust in acting on AI advice (3.20 to 4.00). Participants valued hyperlocal ecological guidance and complementary agent perspectives, while also identifying limits in cultural specificity, ecological grounding, and agent coordination. The work advances relationship-centered AI, offering design implications for multi-agent systems that support food sovereignty, community resilience, and cultural preservation.

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.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces CultivAgents, a relationship-centered multi-agent system for personalized gardening support grounded in ethics of care, with specialized agents for user experience, local environmental conditions, and ethnobotanical knowledge. It reports a three-phase mixed-methods evaluation with domain experts (n=3), HCI researchers (n=7), and community gardeners (n=5), including pre/post surveys showing mean increases in confidence (3.00 to 3.60), motivation (4.00 to 4.40), and trust in AI advice (3.20 to 4.00), plus qualitative feedback on system strengths and limits in cultural specificity and coordination.

Significance. If the reported outcomes hold under stronger controls, the work contributes design knowledge for multi-agent systems in HCI that integrate socio-cultural context, with potential implications for applications supporting well-being, food autonomy, and community resilience. The participatory elements and identified limits offer usable guidance for future relationship-centered AI.

major comments (2)
  1. [Evaluation with community gardeners] Evaluation section (community gardeners cohort): The pre/post mean shifts (confidence 3.00→3.60, motivation 4.00→4.40, trust 3.20→4.00) are reported from n=5 without a control condition, randomization, statistical tests, error bars, or explicit discussion of confounds such as participation effects or social desirability. This directly undermines the abstract's causal claim that CultivAgents 'helped gardeners translate interest into situated action.'
  2. [Abstract and results] Abstract and results presentation: The central empirical claim rests on these untested self-report deltas; the manuscript should either qualify the attribution language or supply additional analysis (e.g., qualitative triangulation or larger sample) to make the evidence proportionate to the stated contribution.
minor comments (1)
  1. [Abstract] The abstract states a 'three-phase mixed-methods study' but does not enumerate the phases; a one-sentence outline would improve clarity for readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's detailed review and constructive suggestions regarding the evaluation design and the strength of the empirical claims. We address each major comment below.

read point-by-point responses

  1. Referee: [Evaluation with community gardeners] Evaluation section (community gardeners cohort): The pre/post mean shifts (confidence 3.00→3.60, motivation 4.00→4.40, trust 3.20→4.00) are reported from n=5 without a control condition, randomization, statistical tests, error bars, or explicit discussion of confounds such as participation effects or social desirability. This directly undermines the abstract's causal claim that CultivAgents 'helped gardeners translate interest into situated action.'

    Authors: We agree that the small sample size and lack of a control condition mean the pre/post shifts cannot support causal claims. The data are descriptive and exploratory. We will revise the abstract, results, and evaluation sections to remove any causal attribution language, explicitly describe the study as exploratory, add error bars to the reported means where the raw data permit, and expand the limitations discussion to address potential confounds including participation effects and social desirability. We will also strengthen the integration of qualitative data from the participatory design activities to provide triangulation. revision: yes

  2. Referee: [Abstract and results] Abstract and results presentation: The central empirical claim rests on these untested self-report deltas; the manuscript should either qualify the attribution language or supply additional analysis (e.g., qualitative triangulation or larger sample) to make the evidence proportionate to the stated contribution.

    Authors: We will qualify the language in the abstract and results to make clear that the reported shifts are preliminary self-report observations from a small exploratory cohort. We will add further qualitative triangulation drawn from the expert interviews and participatory sessions to contextualize the quantitative indicators. A larger controlled study lies outside the scope of the current work; we will note this explicitly as a limitation and direction for future research. revision: yes

Circularity Check

0 steps flagged

No significant circularity: empirical evaluation only

full rationale

The paper reports results from a three-phase mixed-methods study (expert interviews, pre/post surveys, participatory design) with no equations, models, derivations, or predictions. The central claim consists of observed mean shifts in self-reported survey items from n=5 participants. These are direct empirical measurements, not quantities derived from any fitted parameter, ansatz, or self-citation chain that reduces to the input by construction. No load-bearing self-citations, uniqueness theorems, or renamings of known results appear in the provided text. The evaluation is self-contained against external benchmarks (survey responses) and does not invoke any internal derivation that could be circular.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The evaluation depends on standard HCI assumptions about the validity of small-scale mixed-methods studies and self-reported metrics; the three agents are software components introduced by the authors rather than independently evidenced entities.

axioms (1)
  • domain assumption Pre/post survey differences in a small non-controlled sample reliably indicate system effectiveness for relationship-centered AI.
    Invoked in the results section describing the n=5 community gardener outcomes.

pith-pipeline@v0.9.0 · 5799 in / 1303 out tokens · 25120 ms · 2026-05-25T03:55:57.323340+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages · 1 internal anchor

  1. [1]

    Muhammad Awais, Ali Husain Salem Abdulla Alharthi, Amandeep Kumar, Hisham Cholakkal, and Rao Muhammad Anwer. 2025. Agrogpt: Efficient agricul- tural vision-language model with expert tuning. InW ACV. IEEE, 5687–5696

    work page 2025

  2. [2]

    Angels Balaguer, Vinamra Benara, Renato Luiz de Freitas Cunha, Todd Hendry, Daniel Holstein, Jennifer Marsman, Nick Mecklenburg, Sara Malvar, Leonardo O Nunes, Rafael Padilha, et al. 2024. RAG vs Fine-tuning: Pipelines, Tradeoffs, and a Case Study on Agriculture.arXiv:2401.08406(2024)

    work page arXiv 2024

  3. [3]

    Aaron Bangor, Philip T Kortum, and James T Miller. 2008. An empirical evaluation of the system usability scale.Intl. Journal of Human–Computer Interaction24, 6 (2008), 574–594

    work page 2008

  4. [4]

    john Brooke. 1996. SUS: A ’Quick and Dirty’ Usability Scale. InUsability Evalua- tion In Industry. CRC Press. Num Pages: 6

    work page 1996

  5. [5]

    2026.Tiny Gardens Everywhere: The Past, present, and future of the self-provisioning city

    Kate Brown. 2026.Tiny Gardens Everywhere: The Past, present, and future of the self-provisioning city. W.W. Norton & Co

    work page 2026

  6. [6]

    Michelle Chang, Chenyi Shen, Aditi Maheshwari, Andreea Danielescu, and Lining Yao. 2022. Patterns and opportunities for the design of human-plant interaction. InDIS. 925–948

    work page 2022

  7. [7]

    Olivia Doggett, Kelly Bronson, and Robert Soden. 2024. Migrant Farmworkers’ Experiences of Agricultural Technologies. InCHI (CHI ’24). ACM, New York, NY, USA. doi:10.1145/3613904.3642263

    work page doi:10.1145/3613904.3642263 2024

  8. [8]

    Olivia Doggett, Matt Ratto, and Priyank Chandra. 2023. HCI Research on Agricul- ture: Competing Sociotechnical Imaginaries, Definitions, and Opportunities. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI ’23). ACM, New York, NY, USA. doi:10.1145/3544548.3581081

    work page doi:10.1145/3544548.3581081 2023

  9. [9]

    Mahmoud, et al

    Mossad Fanuel, Mahmoud N. Mahmoud, et al . 2025. AgriRegion: Region- Aware Retrieval for High-Fidelity Agricultural Advice.arXiv preprint(2025). arXiv:2512.10114 https://arxiv.org/abs/2512.10114

    work page arXiv 2025

  10. [10]

    Rashmi Gangamma et al . 2024. Role of Gardening in Mental Health, Food Security, and Economic Well-Being in Resettled Refugees: A Mixed Methods Study.Journal on Migration and Human Security12, 1 (2024), 3–18. doi:10.1177/ 23315024231216112

    work page 2024

  11. [11]

    Lissy Goralnik, Lucero Radonic, et al . 2023. Growing Community: Factors of Inclusion for Refugee and Immigrant Urban Gardeners.Land12, 1 (2023), 68. doi:10.3390/land12010068

    work page doi:10.3390/land12010068 2023

  12. [12]

    Taicheng Guo, Xiuying Chen, et al. 2024. Large Language Model based Multi- Agents: A Survey of Progress and Challenges. InIJCAI (IJCAI 2024). 8048–8057. https://arxiv.org/abs/2402.01680 arXiv:2402.01680

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  13. [13]

    Ana O Henriques, Anna RL Carter, Beatriz Severes, Reem Talhouk, Angelika Strohmayer, Ana Cristina Pires, Colin M Gray, Kyle Montague, and Hugo Nicolau

    work page

  14. [14]

    InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems

    A feminist care ethics toolkit for community-based design: Bridging theory and practice. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–26

    work page 2025

  15. [15]

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Steven Yau, Zijuan Lin, Liyang Zhou, et al . 2024. MetaGPT: Meta programming for a multi-agent collaborative framework. In ICLR, Vol. 2024. 23247–23275

    work page 2024

  16. [16]

    Agata Krzyżewska. 2025. AI Advice for Amateur Food Production: Assessing Sustainability of LLM Recommendations.Sustainability17, 23 (2025), 10466. doi:10.3390/su172310466

    work page doi:10.3390/su172310466 2025

  17. [17]

    J R Landis and G G Koch. 1977. The measurement of observer agreement for categorical data.Biometrics33, 1 (March 1977), 159–174

    work page 1977

  18. [18]

    Jason Edward Lewis, Angie Abdilla, Noelani Arista, Kaipulaumakaniolono Baker, Scott Benesiinaabandan, Michelle Brown, Melanie Cheung, Meredith Coleman, Ashley Cordes, Joel Davison, et al . 2020. Indigenous protocol and artificial intelligence position paper. (2020)

    work page 2020

  19. [19]

    Lewis, H¯emi Whaanga, and Caleb Yolgörmez

    Jason E. Lewis, H¯emi Whaanga, and Caleb Yolgörmez. 2024. Abundant Intelli- gences: Placing AI Within Indigenous Knowledge Frameworks.AI & Society40 (2024), 2141–2157. doi:10.1007/s00146-024-02099-4

    work page doi:10.1007/s00146-024-02099-4 2024

  20. [20]

    Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. 2023. Camel: Communicative agents for" mind" exploration of large language model society.NeurIPS36 (2023), 51991–52008

    work page 2023

  21. [21]

    Jinliang Li, Shuaitao Yang, and Delin Chen. 2024. Citizen-Centric Multiagent LLM Systems. InProceedings of the 2024 International Conference on Information Technology for Social Good (GoodIT ’24). ACM, New York, NY, USA. doi:10.1145/ 3677525.3678636

    work page arXiv 2024

  22. [22]

    Susan Loh, Marcus Foth, and Yasu Santo. 2024. The More-than-Human Turn in Human-Plant Interaction Design.International Journal of Human-Computer Studies181 (2024), 103128. doi:10.1016/j.ijhcs.2023.103128

    work page doi:10.1016/j.ijhcs.2023.103128 2024

  23. [23]

    Muhammad Murad, Muhammad Ahmed, Nizam ul din, et al. 2026. Agentic AI Framework to Automate Traditional Farming for Smart Agriculture.AgriEngi- neering8, 1, Article 8 (2026). doi:10.3390/agriengineering8010008

    work page doi:10.3390/agriengineering8010008 2026

  24. [24]

    Vincent Nguyen, Willow Hallgren, Ashley Harkin, Mahesh Prakash, and Sarvnaz Karimi. 2025. My Climate CoPilot: A Question Answering System for Climate Adaptation in Agriculture. InACL. 62–70

    work page 2025

  25. [25]

    Elizabeth Onyango, Destiny Otoadese, et al . 2025. Exploring Neighborhood Transformations and Community Gardens to Meet the Cultural Food Needs of Immigrants and Refugees: A Scoping Review.Health & Place92 (2025), 103433. doi:10.1016/j.healthplace.2025.103433

    work page doi:10.1016/j.healthplace.2025.103433 2025

  26. [26]

    Destiny Otoadese, Issa Kamara, and Elizabeth O. Onyango. 2025. Growing Roots: The Role of Collective Community Gardening in Cultural Food Insecurity and Social Integration of African Immigrants in Alberta, Canada.BMC Public Health 25 (2025), 2271. doi:10.1186/s12889-025-23552-z

    work page doi:10.1186/s12889-025-23552-z 2025

  27. [27]

    Theophilus Lincoln Owiti and Andrew Kiprop Kipkebut. 2025. Enhancing AI- Driven Farming Advisory in Kenya with Efficient RAG Agents via Quantized Fine-Tuned Language Models. InAfricaNLP 2025 Workshop, Constantine Lignos, Idris Abdulmumin, and David Adelani (Eds.). ACL, Vienna, Austria, 24–30. doi:10. 18653/v1/2025.africanlp-1.5

    work page 2025

  28. [28]

    Summers, et al

    Ioana Panţiru, Rachael M. Summers, et al. 2024. The Impact of Gardening on Well- being, Mental Health, and Quality of Life: An Umbrella Review and Meta-analysis. Systematic Reviews13 (2024), 45. doi:10.1186/s13643-024-02457-9

    work page doi:10.1186/s13643-024-02457-9 2024

  29. [29]

    O'Brien and Carrie Jun Cai and Meredith Ringel Morris and Percy Liang and Michael S

    Joon Sung Park, Joseph C. O’Brien, Carrie J. Cai, et al. 2023. Generative Agents: Interactive Simulacra of Human Behavior. InUIST (UIST ’23). ACM, New York, NY, USA. doi:10.1145/3586183.3606763

    work page doi:10.1145/3586183.3606763 2023

  30. [30]

    Maneesha Perera et al. 2025. Indigenous Peoples and Artificial Intelligence: A Systematic Review and Future Directions.Big Data & Society(2025). doi:10.1177/ 20539517251349170

    work page 2025

  31. [31]

    Namita Singh, Jacqueline Wang’ombe, Nereah Okanga, Tetyana Zelenska, Jona Repishti, Sanjeev Mishra, Rajsekar Manokaran, Vineet Singh, Mohammed Irfan Rafiq, Rikin Gandhi, et al. 2024. Farmer. Chat: Scaling AI-Powered agricultural services for smallholder farmers.arXiv:2409.08916(2024)

    work page arXiv 2024

  32. [32]

    Lucie Sovová, Petr Jehlička, and Petr Daněk. 2021. Growing the Beautiful An- thropocene: Ethics of Care in East European Food Gardens.Sustainability13, 9 (2021), 5193. doi:10.3390/su13095193

    work page doi:10.3390/su13095193 2021

  33. [33]

    Reem Talhouk, Kyle Montague, et al. 2022. Refugee Food Insecurity & Technology. Computer Supported Cooperative Work (CSCW)31 (2022), 341–372. doi:10.1007/ s10606-022-09423-w

    work page 2022

  34. [34]

    Samarth Tomar, Samreen Kazi, Vinay Setty, and Avishek Anand. 2024. EcoSage: A Conversational Agent for Garden Management Using RAG. InECIR (ECIR 2024). Springer. https://arxiv.org/abs/2401.06807 arXiv:2401.06807

    work page arXiv 2024

  35. [35]

    Richards, Peter Pypers, Abhishek Ghosh, Jawoo Koo, Shakuntala Johal, and Brian King

    Asaf Tzachor, Medha Devare, Catherine E. Richards, Peter Pypers, Abhishek Ghosh, Jawoo Koo, Shakuntala Johal, and Brian King. 2023. Large Language Models and Agricultural Extension Services.Nature Food4 (2023), 941–948. doi:10.1038/s43016-023-00867-x

    work page doi:10.1038/s43016-023-00867-x 2023

  36. [36]

    Steven Umbrello et al. 2021. Value Sensitive Design to Achieve the UN SDGs with AI: A Case of Elderly Care Robots.Minds and Machines31, 3 (2021), 395–419. doi:10.1007/s11023-021-09561-y

    work page doi:10.1007/s11023-021-09561-y 2021

  37. [37]

    Carolina Villegas-Galaviz and Kirsten Martin. 2023. Moral Distance, AI, and the Ethics of Care.AI & Society39 (2023), 1695–1706. doi:10.1007/s00146-023-01642-z

    work page doi:10.1007/s00146-023-01642-z 2023

  38. [38]

    Yiyang Wang, Chen Chen, Tica Lin, Vishnu Raj, Josh Kimball, Alex Cabral, and Josiah Hester. 2025. Companioncast: A multi-agent conversational ai framework with spatial audio for social co-viewing experiences.arXiv:2512.10918(2025)

    work page arXiv 2025

  39. [39]

    Yiyang Wang, Yiqiao Jin, Alex Cabral, and Josiah Hester. 2026. MASCOT: Towards Multi-Agent Socio-Collaborative Companion Systems.arXiv:2601.14230(2026)

    work page arXiv 2026

  40. [40]

    Wright et al

    Kathrine E. Wright et al . 2021. The Impact That Cultural Food Security Has on Identity and Well-Being in the Second-Generation U.S. American Minority College Students.Food Security13 (2021), 701–715. doi:10.1007/s12571-020- 01140-w

    work page doi:10.1007/s12571-020- 2021

  41. [41]

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, et al. 2024. Autogen: Enabling next-gen LLM applications via multi-agent conversations. InCOLM

    work page 2024

  42. [42]

    Shuting Yang, Zehui Liu, Wolfgang Mayer, Ningpei Ding, Ying Wang, Yu Huang, Pengfei Wu, Wanli Li, Lin Li, Hong-Yu Zhang, et al. 2024. Shizishangpt: An agri- cultural large language model integrating tools and resources. InWISE. Springer, 284–298

    work page 2024

  43. [43]

    Xianjun Yang, Junfeng Gao, Wenxin Xue, and Erik Alexandersson. 2024. PLLaMa: An Open-source Large Language Model for Plant Science.arXiv preprint(2024). arXiv:2401.01600 https://arxiv.org/abs/2401.01600

    work page arXiv 2024