Click on the link below to view the webinar by Cynthia Aghogho from Auburn University, USA, on 4 December 2025 entitled Consumer-Centric Breeding – Extending Tricot as a Participatory Tool for Mapping Cassava Trait Preferences.
Presenter Bio: Dr. Cynthia Idhigu Aghogho is a Postdoctoral Researcher at the Department of Crop, Soil and Environmental Sciences, Auburn University, USA. She is a plant breeder whose research bridges participatory evaluation, sensory science, and predictive breeding. This study was conducted during her time at the International Institute of Tropical Agriculture (IITA), where she worked on integrating consumer insights into cassava improvement.
For more info, contact Cynthia at caghogho@wacci.ug.edu.gh
Summary of presentation: Cynthia showed how the tricot participatory method, originally designed for farmer evaluations, was adapted to map consumer preferences for cassava quality traits. The study compared tricot with conventional hedonic scales to understand how consumers rank cassava genotypes based on colour and texture. The resulting rank data were integrated into predictive modeling, providing a pathway toward consumer-centric and socially responsive cassava breeding that aligns scientific innovation with real-world demand.
| Question | Answer |
|---|---|
| When you say the correlation in the tricot trial is stronger than in the conventional trial and is positive, is it possible to partition that correlation into direct and indirect effects (for example using path or partial correlation analysis), to identify which specific factors are actually driving the observed relationship?” | I didn’t really think about that, but we can, we can do that and add it to our analysis |
| You said that breeders can use this approach earlier in the breeding selection. But one could imagine that you might have19 or 18 genotypes and hundreds of participants to do the evaluation. So when, what is the earliest breeding stage that we can use this approach? | Consumer preference information can be used early in breeding, especially around the advanced yield trial (AYT) stage. By identifying traits (with marker, lab, or spectral data) that are strongly correlated with what consumers like, those traits can be put into a selection index and used as proxy measures to help select better genotypes earlier in the breeding process. |
| Can you talk a little bit about the differences in costs and the duration of the trials? | Tricot is more expensive if we include incentives for consumers, because we used many more people (around 760), and each person got a unique set of three varieties. In contrast, in the classical method about 120 people all evaluated the same eight varieties, so incentives would cost less overall. However, in terms of time, tricot is actually faster and less tiring for consumers. With tricot, they just compare three varieties (A, B, C), but with the classical method they have to score many varieties on a 1–9 scale, and by the time they reach the later varieties they experience fatigue. |
| When working with such a large number of farmers in tricot, how do you organize farmer field days when your tricot plots are many small on-farm trials, rather than one big “mother trial” in a single location? For example, how do you practically bring 500–1000 farmers together and still link that to the smaller tricot plots? . | I’m not involved in on-farm trials or field days that is Bela Teeken from IITA’s work, so you can ask him directly. My work focuses on consumer preference studies, where consumers come to us—usually in a canteen setting—and we give them samples to taste and evaluate. |
| I have a question about your machine learning models. It was not clear to me what you are you predicting. Are you predicting rankings? Are you predicting scores or log words? And what do you use for inputs and outputs? | For tricot, we use the log worth while for the classical, we use the rankings. The inputs are the measurable traits, and outputs are the observer traits. |
| Have you explored whether gari preferred for eba is also preferred for other ways gari is consumed (like “soakings”), or whether different uses might require different quality types? In other words, do quality traits translate across uses, or would certain varieties end up being specialized for specific uses? | We did ask about consumption patterns, and most people reported using gari both for soaking and for eba. We didn’t directly evaluate “soaking quality” in this study, but from previous work we know that a key biophysical trait—the swelling index (how much gari swells when water is added)—is correlated with dry matter. Since soaking quality is related to swelling and dry matter, we assumed that the traits we measured would also be relevant for soaking, and so we treated the overall gari quality traits together. |
| In consumer ranking or scoring, there’s often a halo effect: if people really like one dominant trait of a product, they tend to score all other traits high as well. Did you observe this in your study? And does the “strikeout” method—where you present traits one by one and ask consumers to choose between them—help to reduce this halo effect compared to overall product scoring? | In the classical method, we did see a strong halo effect: when a variety was scored high for one trait (like colour), it was usually scored high for all other traits, because people evaluated everything at once. In contrast, with tricot—where consumers compare just three varieties and choose separately for colour, texture, etc.—that method helps break this pattern and makes their decisions more trait-specific. |