SDG 1: No Poverty
SDG 4: Quality Education
SDG 8: Decent Work and Economic Growth
SDG 10: Reduced Inequality
2. Project Details
Company or Institution
General description of the AI solution
SkillLab’s AI-based solution empowers people to capture their skills, find education and jobs as well as generate tailored job applications. SkillLab makes career guidance accessible to marginalized people and provides a pathway to employment based on a skill-recognition system that is granular, technology-enabled and data-driven.
Users create a skill profile through an AI-based interview that builds on the European Skills, Competences, Qualifications and Occupations framework (ESCO) which contains 13.485 skills and describes 2.942 occupations. The underlying machine learning approach intelligently adapts to the answers of users and won the Google AI Impact award. Our AI allows for quick and responsive skill profiling and allows users to navigate the extensive ESCO framework.
Based on the generated skill profile, SkillLab recommends careers, occupations and, if available, live vacancies in their area. Users can generate job application materials that show their skills, tailored to the requirements of the job. The solution also incorporates local and online education offerings that address the skill gap between the desired job and a user’s skills. SkillLab ensures that the skills taught in education offerings and the skills demanded for vacancies are correctly identified by using a combination of natural language processing and the insights from our data model.
SkillLab helps people access career orientation and guidance. Access to career guidance, labour market information and support in applying for vacancies or choosing education helps individuals to reach their potential, economies to become more efficient and societies to become fairer and more inclusive.
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
In order to guide users in the creation of their skill profile, Skilllab uses an assessment/recommendation engine that intelligently adapts to the answers of users. Given the dataset of 13.485 skills, a survey process or manual use of the database is impossible. The use of AI enables SkillLab to combine the intelligence of an interviewer with a data model that describes all relevant skills in the labor market.
All machine learning components run on Google Cloud Platform and are written in Python. Our system is based on the use of:
– Natural language processing and graph-theory to exploit intrinsic relations within the skill dataset
– Recommender system (hybrid collaborative system model) to intelligently query through the most promising of the 13.485 skills
– Bayesian machine learning to adapt to user-data as user input grows.
– Bayesian optimization algorithms to optimize the balance between the weight in exploiting the given dataset (i.e. ESCO), and the weight to explore possible new trends that are not included in any collected data set
Our technology is mature and proven in its operational environment. The AI has won numerous awards, including the Google AI for Impact challenge, the EUvsVirus Hackathon as well as the Blue Tulip Award for Education and Employment. In the context of these challenges and events, our technology has been exposed to public scrutiny. The validity of our technology is further validated by our official partnership with EURES, the cooperation network formed by public employment services. Finally, we have authored research papers to be published with ETH Zurich for the CEDEFOP and OECD conference on Apprenticeships for Greener Economies and Societies.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
Our Theory of Change is centered around the rationale that offering career guidance and improved skill-based career orientation leads to more sustainable livelihoods by connecting people to occupations, targeted education and upskilling opportunities and, ultimately, careers. You can find our full theory of change and impact framework at https://docs.google.com/spreadsheets/d/1pTYimjz2-Lim5QK6m0q1wExLFS2DsTfhOslTeUKlD8
Our framework relates our performance to the UN’s Sustainable Development Goals and can be measured by the IRIS+ indicators. Our Theory of Change framework focuses on Actions, Outcomes and Impact:
Actions: Delivery of inclusive and skill-based career guidance.
Outcomes: We aim to enable our end users to have a clear understanding of their skill profile and how it relates to potential career opportunities. We measure this by tracking the net promoter score, skill profile fidelity score, number of skill profile output documents generated and percentage of users that identified career goals as well as client satisfaction ratio. Our ambition is that more end users are able to target and participate in education and training and find employment. To this end we track individuals trained, vocational/technical training placement, school/university enrollment and job placements. We also track the number of individuals entering a new experience four weeks after initial skill profile completion. IRIS+ indicators include: Job Placement Rate (PI3527), Individuals Trained (PI2998), and Income (PI9409).
Impact: Sustainable Careers, Improved Livelihoods, Fairer Societies. SkillLab aims to foster sustainable economic participation and long-term financial independence. SkillLab further aims to provide access to those usually unable to rely on career guidance. Directly supported goals include SDG 1, 4, 8 and 10.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
SkillLab incorporated in early 2018. We have spent the past three years conducting intense user testing and developing our solution via alpha and beta stages. We launched in March 2020. In total, over 2.500 users have used our solution post-launch.
SkillLab has been deployed with great success to marginalized communities across diverse economic contexts. Users include refugees, migrants, mature workers, informal workers, victims of gender-based violence and NEET youth. SkillLab helps career counsellors and case workers better understand their beneficiaries and provide better services. Crucially, their beneficiaries significantly improved their ability to identify and describe their skills and capabilities. Users confirm this, reporting that SkillLab helped them identify new skills, gain a deeper understanding of their own skill set and improved their experience during the hiring process. The self-reflective exercise of creating a skill profile and career planning further helped users build self-confidence and motivation, particularly among those furthest removed from the labour market.
Proper collection, use and interpretation of standardized labour market data can play a large role in generating sustained outcomes and anticipating labour market trends. As countries and industries are faced with emerging skills needs and skills mismatches, potential solutions can only come from relying on the language of skills, communicating across stakeholders groups and putting skills at the forefront of active labor market policies, up-skilling, re-skilling and research. SkillLab can enable this transition, using the most widely adopted device, the mobile phone.
One challenge that we face in scaling is the cost of making our solution available in new languages. Integrating new languages is in high demand, but carries significant up-front costs as we have to translate ESCO and continue to update the translations.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
SkillLab takes the ethical implications of AI, trustworthiness and inclusivity very seriously.
SkillLab has very consciously avoided using AI to make any recommendations regarding which careers users should pursue. We merely use AI to aid in capturing the skills of people, jobs and education. With regards to bias in data and algorithms, our AI does not use any personal data, location data or proxy data as input for its skill suggestions. We follow responsible AI guidelines such as the People+AI Guidebook to ensure fair and responsible use. Our AI is trustworthy in all three components throughout the system’s lifecycle and we do our utmost to ensure the accuracy of our engine and minimize any biases.
Technological/algorithmic accuracy of the skill assessment engine means that:
1. The user positively reacts to relevant skills suggested by the assessment engine (Greedy Maximization).
2. The user positively reacts to suggested skills that are not directly obvious or intuitive to humans. This requires our assessment engine to identify and exploit connections between skills outside of conventional thinking (Skill Exploration).
We are developing an entropy measure which captures the above two loss measures as a single, unified scalar value to automatically optimize by this value. Additional qualitative controls are used to monitor the behaviour of the assessment engine.
We have identified key risks. Our skill taxonomy, ESCO, may be biased towards representing the European skills system. By retraining our algorithms with collected user data, we, over time, reduce the inherent bias of ESCO and other taxonomies. We further monitor differences between different user groups (gender, nationality, age, etc.) and their effect on skills selection and other trends in user inputs. We intervene in the algorithm-development when any trends show potential for improvement.