Skip to content

New Deliverable: Report on Commonly Used Algorithms and their Performance

    Good morning all, today our project has reached another milestone by sending another deliverable to the EU. As Hecat intends to build a more ethical algorithm, it is essential that we understand and critique what already exists. This report exists to inform us of the approaches which have already been taken in the algorithm arena, and provides some signposts for which approaches worked best.

    6,536 Algorithm Photos - Free & Royalty-Free Stock Photos from ...

    The report is elegantly written and flows well from one section to another, it begins by outlining and explaining what kind of statistical techniques and models have been utilized to understand the unemployed. This includes all relevant definitions and context, such as a discussion on what profiling is, and how it has been used to achieve various results in the context of unemployment policy.

    The next section deals with decision support systems / artificial intelligence. Computerized systems have long been used to try and select the best candidate for a job, but this is a complex decision and how one ranks the different criteria on offer (education, age, work experience) produces wildly different results. As the authors note, there is also a significant subjective element to this. In each case (i.e. for each job hiring, or post being filled) the employer must indicate which factors they believe to be the most important. In the decision support system models, this will lead to certain attributes or characteristics being listed as more important than others.

    On the one hand there is considerable arbitrariness to this, the decision may be altered by bias or preconceived ideas about what type of person would be good for the role. This may rank a person very poorly because they lack a particular qualification even though they would be ideal for the role. On the other hand those doing the hiring are best positioned to use their knowledge and experience to inform their decision making, and what attributes are required for a person to be successful in the role.

    The following section deals with Labour Flow Networks, and the processes that have been used to try and understand how people move from a state of unemployment to employment and vice versa, but additionally how people flow between different jobs within the same (or similar) industries. Below is a figure taken from the report showing (in part B for example) how enmeshed the employment profile of the various social media and multinational technology companies are, with people flowing from a job in one (i.e. Apple) to elsewhere (most likely Microsoft or Google, but increasingly into Twitter, Uber or Facebook).

    The final section summarizes the methods which have been implemented in the various PES’ (Public Employment Services), which have attempted to find work for the unemployed, and the relative accuracy of these methods. The essential table from this section is reproduced below for your convenience.

    We hope you enjoy reading the report as much as we did, it is informative and provides a window into the world of data science without needing to be a master of statistics (though it wouldn’t hurt!).

    You can find the report on our deliverables page here. Alternatively you can download it via the link below.