Abstract
Altruism, i.e., an act that one does at their own expense that tends to enhance others well-being, is a fundamental human behavior with implications for personal and societal welfare. Hence, modeling altruism is an important building block in designing human-centered computing systems. Traditional methods for understanding an individual's altruistic propensities have been surveys and lab experiments. However, the emerging "personal big data" coming from mobile and ubiquitous devices allows for creation of lower-cost, quicker, automated methods for modeling human behaviors and propensities. We propose a new methodology to model altruism using phone data. Based on analysis of data from a 10-week field study (<inline-formula><tex-math notation="LaTeX">{\text{N}}= {\text{55}}</tex-math> <mml:math> <mml:mrow> <mml:mtext>N</mml:mtext> <mml:mo>=</mml:mo> <mml:mtext>55</mml:mtext> </mml:mrow> </mml:math> <inline-graphic xlink:href="bati-ieq1-2873346.gif"/> </inline-formula> participants), we report that: (1) multiple phone-based features are associated with users' altruistic propensities; (2) phone features-based altruism prediction model yielded significantly better performance than a demography-based model. The results pave way for utilizing "personal big data" to model altruism in multiple commercial and social applications.