User Experience for Artificial Intelligence: An Alexa Skill to Reduce Food Waste.
Conversational user interfaces in private homes are getting more and more common as they are becoming affordable. Still they have many challenges to tackle like the correct parsing of the input into speech or the applying logic. Also it is challenged by a new form of navigation and it has to consider natural language processing which is influencing the users’ behavior. Because of those many technological challenges and the fact that the technology is really new for private households there are not many skills developed by now for the technology. One gap is for example the area of sustainability and the reduce of food waste. Up to 30% of food waste are wasted by consumers. Often the food waste could be reduced if the consumer would have more knowledge about the aspects of food freshness. As the high amount of food waste is leading to further problems like the increase of greenhouse emissions this is an important issue which should be addressed. Even if both aspects are very urgent at the moment, the increasing use of conversational agents and the raising problems of food waste, are not considered together yet. Thereby the main focus of this work is to provide a conversational assistance to support private households to reduce their food waste. In order to simplify the process of food detection this research focuses on the assistance of figuring out the freshness of the fish at home.
Overall the implementation of the alexa skill was conducted in three stages:
● Conducting Interviews
● Interview analysis
● Implementation of skill in dialogue flow
Conducting interviews: Details of the participants are mentioned in the below table. Interviews conducted were semi-structured interviews, The interview guidelines were made prior conducting interviews. The main focus was to get information from participants on approaches of detecting fish freshness and minimizing food wastage.
Interview analysis: After conducting three interviews, transcripts were made for these interviews. Transcripts were analysed using thematic analysis to extract the information and useful insights, which is further used for structuring dialogue flow.
After the first initial research and data collection from experts, an initial dialog flow chart was constructed, subsequently “the first round” of Wizard of Oz method was used for the first time in this study, in this case the Wizard of Oz evaluation method was mainly used to determine and evaluate the accuracy and correctness of the dialog flow to achieve the desired result which is to identify the state of the fish. The first session was carried out i an uncontrolled environment and manner where the participants were instructed to imagine having a fish, and that they can use the system to check the state of the fish.
The final implementation of the skill was carried out with a dialogue flow console of google.
Usage of Wizard of Oz has proven to be a powerful tool in our study, merely because of agility and the simplicity of the process as a whole, as to conduct the evaluation session little effort was needed to implement a system or a fully functioning prototype, a low fidelity prototype that focused on the task being tested and measured was sufficient, and provided a tremendous amount of feedback and information. Another powerful advantage of why the Wizard of Oz method has been reused across the project is its flexibility and adaptability, where with small adjustments or additions to the prototypes or the environment was able to change the evaluation to a completely different aspect of the system. It helped to establish a helpful artificial intelligence voice tool, to support user in reducing their food waste with a main focus on trust and the security of the user.