SAAIL Lab SAAIL Lab
AfriCHI 2025 · ACM Published · ACM AfriCHI 2025

From Art to Algorithms: Co-Designing AI for Clove Grading with Zanzibar's Indigenous Experts

1School of Engineering and Science, Indian Institute of Technology Madras Zanzibar, Tanzania
2Wadhwani School of Data Science and AI (WSDSAI), Indian Institute of Technology Madras, Chennai, India
AfriCHI ACM Conference 2025 (Nov 2025)
From Art to Algorithms: Co-Designing AI for Clove Grading with Zanzibar's Indigenous Experts

Summary: AfriCHI 2025 socio-technical co-design study pairing Zanzibar clove inspectors with segmentation-first mobile AI (99.0% accuracy).

Abstract

We propose a socio-technical co-design approach to modernizing Zanzibar's clove grading ecosystem alongside certified Indigenous grading experts. Technically validating a segmentation-first deep learning pipeline that achieves 99.0% accuracy, we embed inspector domain heuristics, transparent grade rationales, and mobile usability into the interface to foster farmer trust.

Keywords: co-designHCI4Dclove gradingZanzibarAfriCHIindigenous knowledge

Key Findings & Contributions

99.0%
Segmentation-First Acc
Single-clove validation baseline
Co-Design
Participatory HCI
Farmers & ZSTC inspectors
Mpeta Focus
Grade 4 Transparency
Fair farmer pricing compensation
ACM AfriCHI
Published 2025
Human-centered African AI
  1. Paired computer vision validation (99.0% single-clove accuracy via Enhanced U-Net segmentation-first architecture) with participatory HCI co-design involving Zanzibar farmers and ZSTC inspectors.
  2. Identified critical socio-technical asymmetries around manual assessment of Grade 4 (mpeta) fermented cloves that erode smallholder trust.
  3. Formulated an actionable co-design roadmap for low-cost mobile inspection tools amplifying rather than displacing indigenous grading expertise.

Benchmark Results

Socio-Technical Validation & Co-Design Findings (ACM AfriCHI 2025)

DimensionBaseline Observation / FindingSystem Design Requirement
Technical Prototype99.0% accuracy on single-clove gradingSegmentation-first isolation for bulk lots
Quality DeterminerGrade 4 (mpeta) pale-brown fermentation quotaVisual highlight of rejected stems & mpeta
Stakeholder TrustPrice ambiguity at warehouse handoffBilingual transparent visual audit trail on phone

BibTeX

@inproceedings{nyalala2025africhi,
  title={From Art to Algorithms: Co-Designing AI for Clove Grading with Zanzibar's Indigenous Experts},
  author={Nyalala, Innocent and Vincent, Patrick Ndowo},
  booktitle={Proceedings of the African Human-Computer Interaction Conference (AfriCHI 2025)},
  publisher={ACM},
  year={2025}
}
APA Citation

Nyalala, I., & Vincent, P. N. (2025). From Art to Algorithms: Co-Designing AI for Clove Grading with Zanzibar's Indigenous Experts. AfriCHI ACM Conference 2025.