New project to track alcohol in Influencer posts

Artificial intelligence project monitoring influencers’ Instagram posts for alcohol marketing

A new research project led by La Trobe University will use artificial intelligence software to monitor social media influencers’ Instagram posts for alcohol marketing, to inform future regulation and reduce alcohol harms.

Funded by more than $200,000 from the Victorian Health Promotion Foundation (VicHealth), researchers will use deep learning to identify digital images and words featuring alcoholic products.

Lead researcher and Director of La Trobe University’s Centre for Alcohol Policy Research (CAPR), Professor Emmanuel Kuntsche, said the increasing use of social media to market alcoholic products is a major concern.

“There is a demonstrated link between seeing alcohol in the media and increased alcohol craving, impulsive decision-making and hazardous drinking,” said Professor Kuntsche.

“This research intends to enhance our capacity to effectively monitor alcohol exposure and marketing in social media through the development of a deep learning algorithm, which we will make publicly available.”

Professor Kuntsche said the ultimate aim is to inform future policy around this currently poorly regulated area.

“Social media influencers are very persuasive and produce a high return on investment for advertisers,” Professor Kuntsche said.

“Influencer marketing is often fleeting, with little transparency, and is not always declared as alcohol marketing. Thus, effective regulation, enforcement and monitoring of influencer alcohol marketing is a substantial challenge.”

Currently alcohol advertising in Australia is self-regulated and voluntary

According to the Alcohol and Drug Foundation (ADF) the voluntary Advertising Code has ‘four key standards’ of responsible alcohol promotion:

  1. Content cannot target minors or young people under 25 years of age
  2. Content cannot encourage heavy or excessive drinking
  3. Content cannot promote alcohol as a mood enhancer, therapeutic solution, or contributor to success
  4. Content cannot show alcohol being consumed during an activity that requires safety precautions (such as driving or operating heavy machinery).11

The ADF also notes that "popular Australian Instagram influencers have been found to show alcoholic drinks in their posts but not revealed they had been paid by alcohol brands to do so. By using influencers, alcohol companies have found a way to bypass their own voluntary codes and advertise to young people.”

As part of this study, researchers will quantify the kind of alcohol-related content that Instagram influencers post; declared marketing, suspected undeclared marketing, and no marketing.

They’ll systematically identify the marketing characteristics and content in these posts and the way they make declarations of their relationships with the alcohol industry.

The project will also enhance society’s technical capacity to effectively monitor alcohol exposure and (un)declared marketing in social media through the development of a deep learning algorithm, which will be made publicly available free of charge.

CAPR researchers are increasingly using more sophisticated computational technology to develop ways to detect and monitor influence, effects and harm of alcohol.

A recent study showed artificial intelligence could be trained to detect if someone had a blood alcohol content of over 0.05 just by their speech - Deep learning algorithm can hear alcohol in voice.

The current research project is a collaboration between Centre for Alcohol Policy Research at La Trobe University, the Melbourne University Law School, and the George Institute for Global Health.

Media: Courtney Carthy, +61 487 448