A Better Way to Recommend Educational Content

For generations, humans have looked outwards at others for inspiration, individuals of great intellect and accomplishment have been regarded as heroes and model citizens. However, this kind of admiration has not always been shared equally amongst the diverse set of people in the world. As such, with growing awareness, one of the greatest questions of the 21st century is the question of bias, such as gender equality. By investigating user ratings of speakers in TED talks, a popular lecture series that are widely shared for approachable discussions from various experts, we aim to explore whether even experts are affected by bias and whether a platform primarily served to distribute educational content is subject to potential unconscious biases as well. We also aim to explore whether certain types of content are systematically more recommended than others. One form of bias that we are particularly interested is bias in gender perceptions. Previous research in social psychology have found that in various realms and skills, men and women are perceived differently despite lack of difference in actual skill level for example, among STEM and non-STEM students, it was found that there is a signiTicant pro-Male-STEM bias (Farrell et al., 2016). Even among evaluations of teachers , favori t ism towards male professors were found (Boring 2016). It is well documented that recommendation systems are largely correlated with viewership (Zhou et al., 2016). We believe that one of the most important goals of education is to promote a diversity of thinking. By investigating the network structure created by TED talk recommendation lists, we explored how the content is being curated and what kind of content is promoted to users. As stated in their organization page, the goal of TED is to promote a global community, welcoming people from every discipline and culture who seek a deeper understanding of the world. As such, we believe it is important that the recommendation network of TED and similar educational services curate recommendations that promote views towards a diverse set of ideas. Other than gender bias, we hope that by using purely data-driven communitydetection and centrality analyses, we could elucidate the features behind the viewership network and identify the key videos that may be engagement points for TED.com to look at whether talks with certain attributes are systematically emphasized or ignored. From here, we will be able to study the metrics involved that correlate amongst inTluential nodes with hopes to provide metrics that predict future key videos. Continuing, we hope that insight in differences between inTluential and noninTluential nodes could inform more conscious efforts in distributing educational content.