Search

Artificial generation of facial images is increasingly popular, with machine learning achieving photo-realistic results. Yet, there is a concern that the generated images might not fairly represent all demographic groups. We use a state-of-the-art method to generate 10,000 facial images and find that the generated images are skewed towards young people, especially white women. We provide recommendations to reduce demographic bias in artificial image generation.

The competitiveness of the video game market has increased the need for understanding players. We generate player personas from survey data of 15,402 players' 195,158 stated game preferences from 130,495 game titles using the methodology of automatic persona generation. Our purpose is to demonstrate the potential of data-driven personas for segmenting players by their game preferences. The resulting prototype personas provide potential value for game marketing purposes, e.g., targeting gamers with social media advertising, although they can also be used for understanding demographic variation among various game preference patterns.

User perceptions of personas affect the adoption of personas for decision-making in real organizations. To investigate how experience affects the way an individual perceives a persona, we conduct an experimental study with individuals less and more experienced with personas. Quantitative results show that previous experience increases several important perceptions, including willingness to use, empathy, likability, and completeness. Results suggest that methods that increase experience (e.g., training, workshops, scenarios) should be applied alongside persona deployment, as desirable persona perceptions increase with individuals' experience.

Quantitative methods are becoming more common for persona creation, but it is not clear to which extent online data and opaque machine learning algorithms introduce bias at various steps of data-driven persona creation (DDPC) and if these methods violate user rights. In this conceptual analysis, we use Gillespie's framework of algorithmic ethics to analyze DDPC for ethical considerations. We propose five design questions for evaluating the ethics of DDPC. DDPC should demonstrate the diversity of the user base but represent the actual data, be accompanied by explanations of their creation, and mitigate the possibility of unfair decisions.

We address a recommendation task for next likely flight destination to customers of a major international airline company. We compare performance using historical flight data and an actual user evaluation. Using two years of historical flight data consisting of tens of millions of flights, an ensemble and a collaborative filtering approach obtained an accuracy of 47% and 20% using a test set of 100,000 customers, respectively, highlighting the challenge of the domain. We then evaluated our recommendations on 10,000 actual customers, with a 45-45-10 split among ensemble, collaborative filtering, and control group. The overall predictive power employed with real users was 23%, with the ensemble method having a predictive power of 19% and 30% for collaborative filtering. Results indicate that, in complex and shifting domains such as this one, one cannot rely solely on historical data for evaluating the impact of user recommendations. We discuss implications for recommendation systems and future research in this and related domains.