Using artificial intelligence to assist in creating tailored communications and products for disaster preparedness

Back

Introduction

The Whitsundays Climate Change Innovation Hub (Hub) is pleased to be liaising with PhD students to assist their studies of climate change. Work completed during PhD studies, and the results from these studies, will provide important information to help fulfil the Hubs mission to build the resilience of the Whitsundays, and communities everywhere to climate change.

Bio

Elizabeth Forest is a Machine Learning PhD candidate with a Bachelor of Information Technology (Hons). Elizabeth’s honours project used machine learning methods to automatically determine the quality of Minke Whale images. Before beginning her PhD, Elizabeth worked as a software developer for an NQ based start-up that worked on journey management software for safer workplaces. Elizabeth is interested in how complex problems can be solved and simple processes can be automated using technology, with a focus on Machine Learning.

PhD Research Context

Millions of dollars are spent each year due to natural disasters in Australia, with North Queensland (NQ) is a hotspot for cyclones. 92% of NQ residents have experienced at least one cyclone, and the effects of climate change are expected to make cyclones more severe, causing more damage. Therefore, it is important to ensure NQ residents are educated on what they can do to protect themselves and their property. The damage to housing from a cyclone has many effects on the individual, not only the obvious economic impacts, but the personal costs from damage to irreplaceable sentimental items and being displaced from home or work while a dwelling or business premises is repaired.

Risk reduction strategies range from simple, low-cost actions such as tidying a yard or securing loose outdoor items, to more difficult and costly actions such as adding structural upgrades to homes. Understandably, people are more likely to undertake the simple low-cost options, than the more difficult high-cost options, despite the more expensive methods often being more effective.

Effective communication is needed to educate the general public on the benefits of the high-cost options. As an individual’s motivation to perform a protective behaviour is based on their experiences and beliefs, targeted messaging is required for effective communication. A previous study developed a set of distinct groups, or ‘personas’, based on psychological factors. These personas could then be used by a range of people to customize communication. However, the persona creation process can be quite resource intensive. Determining a more efficient method of persona development is necessary to allow personas to be used more often and be applied to a wider range of issues.

Central Question

The central question of this study is can artificial intelligence be used to effectively replicate the decision making of experts for tailoring disaster communication? The project aims to automate the persona development process allowing for a quick and repeatable method of developing personas. This project will apply a variety of state-of-the-art machine learning approaches to the problem of automated persona development for targeted communication, to determine which approach best replicates an expert decision making.

Determining a more efficient method of persona developing is necessary to allow personas to be used more often to effectively customise communication and to be applied to a wider range of issues. Being able to inform the general population in a manner that they are more likely to respond to is hypothesised to result in an increase in the performance of mitigation strategies. Thus, resulting in a safer North Queensland which will take less damage during the next natural disaster.