The world of employment is continuously shifting under the feet of workers. Our report, Mapping Career Causeways: Supporting workers at risk, provides a machine learning powered map to help people navigate this ever-changing landscape.

Using an algorithm, the map helps workers whose roles are at risk from automation, COVID-19 and other factors to identify other, safer roles based on their skills and experience. Now, we are open-sourcing the code and data used to create it, to make sure that it is as accessible as possible.

Although workers are powerfully impacted by the labour market, they often have very little influence over the factors that change it. As some sectors of employment grow, others decline and can disappear altogether, leaving people and communities stranded. The map we have produced charts the possible pathways between 1,600 jobs, based on the skills and work activities involved, as well as factors such as risk of automation, impact from COVID-19 and earning potential. In this way, it is able to recommend transitions from one occupation into other safer roles.

To achieve this, we have brought together multiple datasets including occupational frameworks from ESCO and O*NET, labour force statistics and automation risk estimates, and processed them with multiple data science techniques such as machine learning and natural language processing. All of the results and steps involved are now publicly available via the project’s GitHub repository, linked below.

Producing the map and the research report was just one goal of this project. We are now open-sourcing the code and data to ensure that the work can be explored, reproduced and reused by others.

These resources will primarily be useful for developers, analysts and researchers who are looking to integrate the code or results into their platform or research, or for those who want to examine the report findings in more depth. We believe that there is much more potential in this work to create new insights and augment future digital tools for careers services. 

The map relies on experimental and statistical data science methods as well as decisions made within the context of this project. In opening up the codebase, we have taken two important considerations into account. The first is to include sufficient documentation in both the repository and report to guide users through our choices and understand why certain results have been obtained.

The second is that we have carried out an assessment of the map using crowdsourcing methodology. In doing this, we have been able to provide feasibility ratings for the occupation transitions recommended by the algorithm. This means that we have been able to remove transitions judged to be unrealistic by the general public, and that the remaining transitions can be filtered and examined using this metric. A forthcoming blog post will describe this approach in more detail.

The innovative nature of this work is a reason in itself to make the code open source. Transparency allows others to offer feedback, perform further validation and iterate upon the project.

• Code to reproduce the algorithm and the report findings;
• Datasets including a table of recommended transitions with crowd feasibility ratings;
• Code to create and export the full set of recommended transitions;
• Tutorial notebooks to explore occupation transitions.

We would be very interested to hear about all ways that the code and results in this repository are used, so if you find them useful please get in touch with us at [email protected]. We are also interested in partnering and collaborating with organisations to further extend this work and maximise its reach.

In addition to the report and code, we will be publishing three user guides for the map and its applications later this year. These will contain further practical insights and use cases and will be for non-technical audiences, including those working in policy, employers and the careers services sector.