Algorithm Powered Recruitment Advertising Service

Algorithmic systems in recruitment are already in widespread use in one form or another (Rab-Kettler & Lehnervp, 2019). While for some, they present an immense potential for both employers and jobseekers by offering to correct the bias of the traditional recruitment methods, for others, these algorithmic technologies simply exacerbate many existing biases. The main source of bias found in these AI systems lies in the big data sets that are used in training of the machine learning algorithms (Krishnamurthy, 2019).

Despite their shortcomings, data driven practices and technologies are increasingly relied upon to make decisions about social, economic, and political issues (Richardson, Schultz & Crawford, 2019), but with 'little scrutiny of the assumptions and bias built into' them (Leonelli, 2019) or understanding how they affect the existing lines of accountability (Barocas, Hood & Ziewitz, 2013). In light of these issues, the growing academic field of critical data and algorithm studies are calling for a greater scrutiny of the methods of data creation (Richardson, Schultz & Crawford, 2019) and investment into data governance (Leonelli, 2019). It is within this context that this ethnographic research aims to critically examine the design and use of algorithmic systems focusing specifically on the algorithm powered services that operate in the area of recruitment.

Main objectives of this research:

1. To conduct an in-depth ethnographic study to explore the in-built biases within algorithmic systems designed for recruitment advertising purposes.

2. Using the academic recruitment advertising organisation jobs.ac.uk as an illustrative example, the proposed research will aim to reveal 'intervention points' and ways to create a 'desired bias' which address the lack of diversity.

Research questions

The overarching question explored here is: If the data that feeds the algorithmic systems is biased, what kinds of interventions might there be to overcome them?

To answer this overall research question, these further sub-questions will be explored:

1. How is the data collected, stored and used? What are the qualities of such data?

2. What specific challenges does the process of creation of algorithmic systems face in light of such data?

3. Who is involved in this process and where are the intervention points?

4. To what extent can the structures or road maps be created to guide the appropriate of such technologies?

Research design and Methods

The proposed research project will address these questions in the context of recruitment with an online jobs board as the centre point for data collection. The online jobs board is an algorithm powered service provider that specialises in connecting millions of jobseekers with tens of thousands of advertised jobs. Its digital nature and underlying algorithmic systems have meant that the platform has been able to collect, store and use vast amounts of data.

In order to answer the above research questions, the research will take a perspective that the algorithm-based system is a socio-technical concept (Elish & boyd, 2017) and that an understanding of these technologies 'requires their socio-technical assemblage to be examined' (Kitchin, 2014b). The approach that will be used to unpack the full socio-technical assemblage of them is the one proposed by Kitchin (2014b), that is an ethnographic approach which also includes semi-structured interviews with staff who are involved, directly or indirectly, in the production of its algorithmic systems.

Overall, this project is timely: whilst algorithms are increasingly being researched, the area of recruitment is seldom examined and yet it is so important in shaping the livelihoods of millions of individuals and the net productivity of small and large scale firms.