Product rating: sustainability index and preferences
This protocol is extracted from research article:
How value-sensitive design can empower sustainable consumption
R Soc Open Sci, Jan 13, 2021; DOI: 10.1098/rsos.201418

Sustainability index. It quantifies the support or opposition of a preference by a set of product characteristics found in a product. This support or opposition is compared to a product, existing or hypothetical (‘reference product’ in figure 5), that has all possible characteristics that can support or oppose respectively a preference. Figure 5 illustrates the involved calculations. The sustainability index between a product and a preference is measured using the normalized association scores aggregated over the connected product and preference tags. The normalized aggregated association score of a product-preference pair is the normalized aggregated association score of a product averaged over all preference tags assigned to the preference (Calculation 4 in figure 5). Each normalized aggregated association score between a product and a preference tag (Calculation 3 in figure 5) is calculated by the aggregated association scores of the product tags assigned to this product (Calculation 2 in figure 5) divided by the maximum association score between the reference product and the preference tags of the preference (Calculation 1 in figure 5).

Insights on sustainability of production. Note that by calculating for each preference the density of the sustainability index over all products, new opportunities arise to reason about the following: (i) the sustainability profile of different retailers, (ii) new ways (preferences) with which producers can improve products with a more sustainable profile, and (iii) market gaps where new business ecosystems can evolve with stronger involvement of producers and consumers to accelerate sustainable consumption. For instance, the densities in electronic supplementary material, figure S.10 confirm the more sustainable profile of Retailer B products across the preferences, e.g. higher sustainability index for animal rights, fair trade, recyclability and green farming. Improvements can be made by either introducing new products with better sustainability footprint over these preferences or by improving the existing production practices of the available products.

Product rating. Note that the sustainability index does not require any personal information for its calculation. It only relies on the information of the sustainability ontology, i.e. primitive concepts and tags, that we make available as public-good knowledge. As such, it can be calculated in public computational infrastructure, i.e. servers, public clouds, etc. By contrast, the calculation of the product rating requires personalization with consumers’ preference choices that remain by design locally on their smart phones to protect privacy and limit manipulative nudging. As a result, the product rating is calculated on consumers’ smart phones using the sustainability index values retrieved remotely using a distributed protocol of message passing between smart phones and a project server. The calculation is performed on-demand by consumers when they navigate in the retailer shop and request the rating of the products that are in their close proximity. For each product, the rating algorithm multiplies the sustainability index with the degree of opposition or support of each preference, measured by the distance (offset) from the median preference score (5, remaining neutral). These calculations are summed up and divided by the sum of all distances from the median preference scores. Electronic supplementary material, §SM.2 outlines in more detail the product rating calculation and its computational complexity. The (unscaled) product rating calculation is summarized as follows:

where the product rating values can be scaled to match different grading systems of different countries ([0, 10] in the field tests as supported in electronic supplementary material, § SM.2).

Explainability. Two levels of rating explainability are provided to consumers: (i) product tags and (ii) preferences. Consumers can learn about how each product characteristic influences the rating value by solving equation (34) of electronic supplementary material for a certain product tag, given that all other variables are known. Similarly, consumers can know how each offset of their preferences contributes to the product rating by solving equation (31) of the electronic supplementary material for a certain preference offset.

Preferences selection. The selection of preferences was made on the basis of providing a broad spectrum of different sustainability indicators with which consumers can express their preferences. However, this spectrum is not too broad to the extent of creating a cognitive overload for consumers and lack of comprehension about which preferences influence the rating of products and why. This is critical for the effectiveness of the rating explainability. Moreover, a lower number of preferences decreases the computational cost of the rating algorithm and improves the usability of the smart phone app (see electronic supplementary material, § SM.2.4). This balance is a result of the following process: (i) participation of several stakeholders in the ASSET project meetings and workshops providing insights about how grocery product choices influence different sustainability criteria, (ii) preliminary living laboratory experiments and smaller-scale field tests for feedback acquisition on the preferences, and (iii) choice based on available data, i.e. product and preference tags. Preferences with very similar preference tags or very few preference tags are removed or merged. The set of the final preferences shows a balance between several individual criteria on health (13) versus collective criteria in environment, social and quality aspects (12). The consumers’ feedback during the preliminary tests also determined the strict preferences of gluten-free, lactose-free, vegan and vegetarian products. Fully supporting such preferences results in excluding products that do not fully satisfy them even though they may satisfy other (non-strict) consumer preferences. In other words, strict preferences cancel association scores with other preferences.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.