Consumption of cancer causing products (CCPs) is the root cause of many of the cancers and chronic diseases we aim to cure with translational medicine. Tobacco and alcohol consumption is a function of highly successful, coordinated global business models that promote consumption. The marketing of CCPs can be seen as a disease vector that drives CCP consumption and worthy of surveillance. The present project addresses the marketing of tobacco and alcohol to young people in the current media environment, in which youths increasingly stream their entertainment in a relatively commercial free environment. To address decreasing reach of traditional advertising, companies have sought to embed their brands in the entertainment itself. Alcohol companies actively seek entertainment placement deals. While cigarette companies have agreed not to pay for product placement, they may evade current restrictions by burying placements in the massive streamed entertainment environment. E-cigarette companies have no such restrictions and advertise on television. Youths obtain exposure to brands that spend more on advertising (e.g., Budweiser); these are also the most commonly consumed brands by youth. Product placement exposures may put adolescents at greater risk for CCP consumption. This study represents the first step in assessing that risk--to obtain infoveillance on CCP brand prevalence in streamed entertainment media.
Aim 1 involves a partnership with machine learning researchers at the Jet Propulsion Laboratory (JPL) to develop and validate a computer recognition system for major cigarette, e-cigarette, and alcohol brands in streamed entertainment media.
Aim 1 relies on an existing, large Dartmouth media training library that includes detailed timing for tobacco and alcohol brands from over 2000 contemporary movies. JPL researchers will use the Dartmouth media library, combined with a large corpus of brand logos retrieved from the web, to train and validate an automated Video Content Coding (VCC) system that identifies content associated with key alcohol and tobacco brands. The VCC system will leverage recent techniques to construct robust object detection systems using Convolutional Neural Networks built on large image databases, along with text extraction methods to identify brand names in advertising logos. Second, we will deploy the automated recognition system on the Dartmouth computer cluster to assess the frequency of major cigarette, e-cigarette, and alcohol brand placements from a large sample of streamed/cable and movie entertainment. This proposal involves a unique multidisciplinary team of Dartmouth behavioral scientists and NASA machine learning scientists who will leverage new technologies to study a significant CCP marketing platform. The novel approach offers unprecedented opportunities to conduct surveillance on CCP marketing in large media samples. Our project will set the stage for future research on exposure to this type of marketing and its relation to CCP consumption in youths and lead to a better understanding of the risks CCP brand placements pose to health.
This project addresses the potential exposure of youth to marketing of cancer causing products? tobacco and alcohol? in the popular commercial-free streaming environment, where products are embedded in the media itself through product placement. We will employ machine learning to develop an automated visual recognition system for major tobacco and alcohol brands and deploy that system to evaluate brand placement in a large sample of streamed entertainment from cable (e.g., HBO) and other (e.g., Netflix Originals) companies. This surveillance program will have the capability to monitor embedded marketing across many platforms and ultimately allow us to assess youth exposure to embedded brands and its relation with early onset of tobacco and alcohol use.
