PLOS One / Aggregation processes in customer rating systems : insights from an economic decision experiment / Dirk van Straaten, Behnud Mir Djawadi , Vitalik Melnikov, Eyke Hüllermeier, René Fahr. In: PLOS One. Paderborn : Universitätsbibliothek, 2026
Inhalt
- Aggregation processes in reputation systems
- Aggregation metrics background
- Derivation of aggregation functions
- Arithmetic mean (AM). We commence our analysis with the arithmetic mean where the valence score is calculated by weighting the rating categories in proportion to their scale values. The intuitive interpretation, ease of computation, and documented correla
- Focus on the highest category (FIV). As an alternative to the arithmetic mean, our first approach posits that customers calculate a product’s valence by prioritizing five-star reviews. This assumption is grounded in the prevalence of J-shaped rating distr
- Focus on the lowest category (ONE). Likewise customers may put emphasis on the lowest category to capture the product’s valence. Empirical evidence from [21] demonstrates a shift from unidimensional to multidimensional reputation systems, revealing that t
- Binary perception of ratings (BIN). [23] provide evidence for a cognitive simplification pattern they term the binary bias in consumer evaluation of rating distributions. Their findings indicate that individuals tend to categorize ratings in a dichotomous
- Focus on positive ratings (POS). Empirical evidence suggests that 5-star ratings may not always serve as a reliable indicator of true product quality, as they can be subject to various cognitive biases and strategic manipulation (e.g., [24,25]). Consequen
- Focus on negative ratings (NEG). Similarly, there is growing evidence that consumers who place heightened emphasis on negative reviews may not rely exclusively on 1-star ratings, particularly due to skepticism regarding the authenticity of online feedback
- Median (MED). As a final aggregation function, we consider the median, which is defined as the central value in an ordered ranking of all observed ratings. Specifically, the ratings are sorted in ascending order, and the median corresponds to the category
- Experimental design
- Data and statistical model
- Results
- Descriptive statistics
- Aggregation behavior derived from statistical model
- Heterogeneity in aggregating product information distributions
- Impact of numerical information
- Conclusion
- Supporting information
- References