Understanding Targeted Video-Ads in Children's Content

As the volume of online video entertainment via streaming increases, ever so more are users targeted by online advertisement algorithms. Nevertheless, this rise in targeting and revenue does not come without any concerns. That is, even though the online advertising business model has is very successful, nowadays, rising societal concerns regarding the ethics and extent to which such algorithms agree with the laws of different countries are also present. Motivated by the dichotomy above, we here explore how targeted video-ads meet the regulatory policies regarding children advertising in Brazil and Canada. To perform our study, we create synthetic user personas that watch YouTube videos daily. Our personas are tailored to stream children's content while controlling for several variables (e.g., gender, country, and type of content streamed). With the data gathered, our analyses reveal statistical evidence of algorithmic targeting in videos geared towards children. Also, some of the advertised products (e.g., alcoholic beverages and fast-food) go directly against the regulations of the studied countries. With advertisements being matched to users by machine learning algorithms, it is impossible to state whether regulations are not followed on purpose (e.g., advertisers gaming the system). Nevertheless, our findings and discussion do raise a flag that regulations may not be sufficient, and content providers may still need to audit systems to meet the regulations.