Project Slenky is targeting young audiences (13-24 year olds) with completely different social and business agenda. Instead of exploiting personal data, we want to use it for the benefit of our users, creating content which engages, inspires and includes and makes profit from companies who want to invest into local communities and their public relations.
Together with our new partners for Slenky, (Aston University) we are creating Ethical and Transparent AI systems which would further benefit this young generation feel more included and invited in our society.
There are three groups of stakeholders on Slenky:
- A variety of young people looking for inspiration and opportunities. They come from all walks of life from socially and economically engaged people to disconnected youth.
- SME’s, corporations, education and government organisations who want to provide opportunities for young people
- Slenky as a business that has its specific business goals and various KPIs to measure its success.
What makes Slenky different?
Typically, major social networks want to maximise the time you spend on the platform. Their recommendation engines sometimes use clickbites. Data is typically used for targeted advertising which is not exactly servicing users. It might offer: fake news, misinformation, selling products to their users. We do not share our data with any 3rd party providers. We do not want to use it for purposes similar to other social platforms.
Instead, we want to know what drives them, what motivates, and what inspires them. We will build a psychological profile of our users based on our data. It will be used to build a different type of recommendation engine. We will recommend shots and opportunities to each user individually, so they can find their shot fast, spending the least amount of time on the platform. No misinformation or flogging stuff our users do not need.
The order of presentation of the data is important because people are not all the same. Someone will find the information of location and the time of the event most important, while someone else will find the content and the activities important or the information about the company that is providing it. After our new recommendation engine shows the shot, the shot details need to be presented in the personalized way. Using various data collection and AI, we will learn how each users likes the information to be presented to them, increasing the chance for them to like and take this shot.
Innovation about user profiles
We anticipate to collect sufficient data to enable the creation of detailed user profiles from a wide range of sources. After several deliberations with emphasis on the nature of the Slenky application and the anticipated behaviour of the intended users, the consortium has agreed that using surveys for collecting personal views is likely to fail. Instead, we expect to determine behavioural patterns from observation of interface actions (e.g., browsing delays, response time, repetitive swipes), interests expressed towards certain opportunities (e.g., associations between opportunities and perceived benefits). Knowledge, background and skills (e.g., reflections on own aptitudes), goals (e.g., making relations between concepts and opportunity rewards), interaction preferences, individual characteristics, group profiles and contextual information based on both individual backgrounds and available opportunities.
It is anticipated that user input and feedback will be continuous in a number of ways as users can express their preferred opportunities, rate the description of opportunities provided by the application, highlight terms deemed confusing and selects the level of own confidence when selecting opportunities in relation to the clarity of the description. We will consider how intelligent user profiling techniques such as Bayesian networks, association rules, case-based reasoning, as well as genetic algorithms, and classification techniques (fuzzy logic) contribute in the development of intelligent user profiling that accurately represents individual characteristics and requirements. As part of the project’s outputs, we will deliver reliable user psychological profiling tools.
Innovation about language
Analytics and recommendation – we expect to investigate whether machine learning algorithms can be trained to provide a sufficient level of ‘translation’ between the terms used by recruiters in advertising certain opportunities and the language used by users when describing their skills and capabilities, as well as the searches they conduct when looking for jobs. The team iscapable to examine alternatives such as supervised and unsupervised learning, as well as semi-supervised learning and reinforcement learning. Emphasis will be given on how Natural Language Processing (NLP) can be used to assess the use of linguistic components and whether sufficient patterns emerge to enable the use of certain terms interchangeably. We will deliver language analytical tools, measuring the KPIs of the linguistic components, recommending alternatives in real time during content creation, and delivering modified texts/text options in real time to match each individual user’s language profile.
Innovation about recommendation engine
We will focus on the development of recommender systems that may combine collaborative filtering (i.e. how users have interacted with opportunities via Slenky in the past) and content-based systems (i.e. using information available from user profiling). We aim at delivering a content recommendation engine powered by ML and AI, capable of analysingplatform content, matching it to users and constantly improving the matching KPIs by monitoring user reactions. We will also deliver a content presentation engine, powered by ML and AI, analysinguser needs (via A/B testing, feedback etc.) and being able to do this in real time for each individual user. Throughout the project, we will also conduct thorough risk management on both the technical and ethical sides.
