Learning by Collaborative and Individual‐Based Recommendation Agents

Intelligent recommendation systems can be based on 2 basic principles: collaborative filters and individual‐based agents. In this work we examine the learning function that results from these 2 general types of learning‐smart agents. There has been significant work on the predictive properties of each type, but no work has examined the patterns in their learning from feedback over repeated trials. Using simulations, we create clusters of “consumers” with heterogeneous utility functions and errorful reservation utility thresholds. The consumers go shopping with one of the designated smart agents, receive recommendations from the agents, and purchase products they like and reject ones they do not. Based on the purchase–no purchase behavior of the consumers, agents learn about the consumers and potentially improve the quality of their recommendations. We characterize learning curves by modified exponential functions with an intercept for percentage of recommendations accepted at Trial 0, an asymptotic rate of recommendation acceptance, and a rate at which learning moves from intercept to asymptote. We compare the learning of a baseline random recommendation agent, an individual‐based logistic‐regression agent, and two types of collaborative filters that rely on K‐mean clustering (popular in most commercial applications) and nearest‐neighbor algorithms. Compared to the collaborative filtering agents, the individual agent (a) learns more slowly, initially, but performs better in the long run when the environment is stable; (b) is less negatively affected by permanent changes in the consumer's utility function; and (c) is less adversely affected by error in the reservation threshold to which consumers compare a recommended product's utility. The K‐mean agent reaches a lower asymptote but approaches it faster, reflecting a surprising stickiness of target classifications after feedback from recommendations made under initial (incorrect) hypotheses.     

Differential partitioning of extended experiences

This article focuses on the effect of the perceived cohesiveness of experiences, whether composed of single or multiple parts, on their overall hedonic evaluations. Four experiments demonstrate the effects of partitioning on decision makers’ evaluation of extended experiences. First, patterns (i.e., improving vs. deteriorating trends) strongly influence how experiences are evaluated. Second, increased partitioning of an experience reduces the effect of the overall trend and results in more equal weighting of its parts. Third, breaking experiences at strategic points (i.e., local maxima and minima) influences the overall evaluation of experiences as well as the prediction of their future levels. These results suggest that components of sequences are evaluated similarly to the way whole sequences are evaluated and that experiences composed of multiple components are evaluated relatively more on the basis of their individual intensity and less based on their overall pattern.     

Pornography,preference for porn‐like sex,masturbation, and men's sexual and relationship satisfaction

Pornography use, preference for “porn‐like” sex, masturbation, and sexual and relationship satisfaction were assessed among two samples of men (N_{Study 1} = 326, N_{Study 2} = 335). Frequent pornography use was associated with sexual dissatisfaction, greater preference for porn‐like sex, and more frequent masturbation in both studies. Pornography use was associated with relationship dissatisfaction in Study 2 only. The data did not support the notion that pornography negatively impacts sexual or relationship satisfaction via preference for porn‐like sex. In fact, it may bolster sexual satisfaction by promoting sexual variety. The data were consistent with a model in which pornography negatively, indirectly affects sexual and relationship satisfaction via masturbation frequency. Pornography use may have multiple opposing influences on sexual satisfaction.     

Shape‐ and Trait‐Congruency: Using Appearance‐based Cues as a Basis for Product Recommendations

This research demonstrates that a consumer's physical appearance—and, more specifically, his or her body size—predictably influences the product(s) that the consumer is recommended. Four studies conducted in both field and lab settings show that agents more frequently recommend round (vs. angular) shaped products to heavier targets, notably for products and categories in which body size is irrelevant (e.g., lamps and perfume). We attribute this to a combination of shape‐congruency and trait‐congruency, whereby individuals choose products for others based on shared dimensions of the person and product.     

A randomised trial of the effect of appearance versus health-framed messages on engagement with an online brief screening and alcohol reduction intervention

Objectives: To compare the impact of appearance versus health-framed messages on engagement in a brief web-based risk screening and alcohol reduction intervention.

Design: Randomised trial delivered via Drinkaware’s website. Visitors were exposed to appearance (n?=?51,588) or health-framed messages (n?=?52,639) directing them towards an AUDIT-C risk screening questionnaire. Users completing this questionnaire were given feedback on their risk level and extended frame-congruent information.

Outcomes: The primary outcome is completion of the AUDIT-C questionnaire. The secondary outcome is whether the participant accessed any of four further resources.

Results: The appearance-framed message led to a small but significant increase in the number of users completing the AUDIT-C compared to the health-framed message (n?=?3,537, 6.86% versus n?=?3,355, 6.37%, p?<?0.01). Conversely, following subsequent risk feedback, users exposed to extended health-framed information were more likely to access further resources (n?=?1,146, 2.17% versus n?=?942, 1.83%, p?<?0.01).

Conclusions: Physical appearance-framed messages increased the likelihood of engagement with an online alcohol screening and brief intervention tool, whereas health-framed messages increased the likelihood of accessing further resources. This highlights the potential for the use of multi-level approaches in alcohol reduction interventions.