Coke bottles by drew-taylor

Codethecity 13 – Hack society’s relationship with Alcohol

Introduction

This event, which was sponsored by the Public Health team at NHS Grampian took place on 19-20 May 2018.

The concept was simple – how do we improve society’s relationship with alcohol.

With 21 people there on the Saturday morning we got stuck in.

Challenges

The room came up with over 20 challenges that we could tackle. Everyone identified through voting which they would like to work on, then were asked whether they were prepared to champion one challenge. Three teams emerged – each with enough people to get started on identifying the issues – and coming up with some ideas.

The Teams

The three teams were Recovery Squirrels, CantchDontchaWontcha, and Mindfulness. Each team was assisted wonderfully by a young artist who produced superb logos for them!

Recovery Squirrels logo
Recovery Squirrels logoRecovery Squirrels

Recovery Squirrels

This team worked on ideas to assist those in recovery make informed choices when interacting with support agencies, providers etc through the creation of a code or set of values.

Their first update is available here on Periscope as is their final update from Sunday afternoon.

CantchDontchaWontcha

This team focussed on the design and marketing of alcoholic drinks, cans and bottles. Not only is it difficult to tell from a can’s front whether it is, or is not alcoholic, but there appears to be a trend in marketing craft beers to move to the cartoon-ish, colourful designs which are potentially misleading.

cantcha design
cantcha logo

So, we put together a take on the ‘captcha’ – the ‘cantcha’ – where nine random designs are shown – and the site user has to correctly identify the alcoholic drinks. It is harder than it sounds. That led to the team name.

You can view a video of the mid- Saturday afternoon update on periscope.

We also worked on identifying and classifying 238 drinks cans / bottles. How colourful the design was, how playful the fonts, how much it relied on cartoon characters, and how clearly informative the labelling was.

Whiteboard of pseudo-data
Whiteboard of pseudo-data

Using that information we created some R code to created some machine learning algorithms to try to predict from the label attributes whether a) it was alcoholic or not, and whether a drinker would perceive it as alcoholic or not.

This is a video of the final update  of the weekend.

There is much we could do to build on this with time and some investment:

  • Using publically available machine learning tools such as Google Vision to automatically extract and score assessments for colourfulness, cartoons, faces looking under 25 % space allocated to saying the product is alcoholic, etc.
  • Independently contrast the findings against a panel of school children to label the images as appealing/ not appealing to train the model further.
  • Develop a CAPTCHA device. These are normally used to differentiate between bots and humans by asking the user to identify the images that have certain features. Changing this to drinks that are alcoholic or non-alcoholic would provide data to further refine the model.
  • Develop the model to include the full range of Portman Group CAP rules.
  • Build a bot to automate the process of highlighting problematic products and bringing it to an authority’s attention.

Mindfulness

Mindfullness logo
Mindfullness logo

This team focussed on where people in recovery – or those who did not want to drink alcohol could go for a night out in Aberdeen. Where was a safe space? Where did not serve alcoholic drinks.

Their first update video was published here.

This focus narrowed onto where could they drink mocktails – to how could we encourage the restaurant trade to serve cocktails – and eventually to the concept of MocktailsAwards to encourage the provision of high quality non-alcoholic drinks.

Mocktails Awards
Mocktails Awards

In addition to some very impressive marketing materials and graphics, they even had time to create a new Twitter account and a Medium post about their project.

Their final update is also available.

Conclusion

This was a very productive event – where three teams worked well together, despite being formed on the Saturday morning. They were focussed on tasks – and providing prototype solutions to those. In that they all scucceded. It will be interesting to see whether some of these are taken forward beyond the bounds of the weekend, as the best CTC outcome projects are.

And, as is the case with our weekends, we pushed the teams to use Github to share their work – generating open data and open code. You can find a link to the three repos here.