#data – BOS

Sustainable labour market integration: challenges and advancements in algorithmic profiling of jobseekers

By Clément Brébion and Janine Leschke

◦ 5 min read ◦

The number of countries that are using algorithms to profile jobseekers has been on the rise since the 1990s. Algorithmic profiling aims at identifying individuals with little counselling needs, and those for whom intensive counselling and active labour market policies (ALMP) are expected to have the largest returns. The ultimate goal is to target services and thereby expenditures towards the latter. In a dual context of budget constraints and of technological innovations (which makes it possible to build and analyse large register databases), profiling algorithms are increasingly seen as an important vehicle to identify and target those unemployed who are most likely to become long-term unemployed. In an EU-funded project, HECAT – Disruptive Technology Supporting Labour Market Decision Making, we question this consensus. The goal of the project is to go beyond state-of-the-art profiling tools and develop a tool that will allow jobseekers and counsellors to get a snapshot of their labour market situation and a better sense of their labour market options.

State-of-the-art statistical profiling tools carry important shortcomings. One of them relates to the outcome category when used for defining the profiling categories. Most profiling algorithms approach jobseekers’ needs for counselling and for training programs by measuring their likelihood to remain unemployed for more than 6 or sometimes, 12 months. Usually, any type and length of employment spell is counted as a successful exit from unemployment in these models. Research on the causes and consequences of long-term unemployment (LTU) is extensive and we know that an early identification of the jobseekers that are likely to fall into LTU to take action at the earliest stage possible is key.

However, the mere focus on exits towards any type of employment is problematic. On individual grounds first, it disregards the agency of the unemployed by ignoring her lived experience of unemployment and wishes and aspirations for future labour market integration. Second, such a focus on exits without job quality in focus, can also be dysfunctional and inefficient both from the perspective of the individual and the PES as unsustainable labour market integration is likely to lead to vicious circles where people circle between (short-term) employment and unemployment.

In order to address this shortcoming, in deliverable 2.1 of the HECAT project, we discuss the scope for using job quality information in profiling and job matching tools. We develop a list of 24 items covering 7 dimensions that we see important to take into account to meet SDG (sustainable development goal) 8 on decent jobs and economic growth [1]. We do so by drawing on established job quality indices (e.g. here and here).

By putting the quality of jobs in focus, such an approach provides a more complete and sustainable vision of the labour market to the unemployed and the job counsellors and thereby increase their agency.

As we outline in the deliverable there are a number of challenges with this approach. This includes the high complexity of multi-dimensional job quality indices in view of an efficient and usable counselling and visualisation tool as well as a lack of sufficiently detailed job quality indicators on the level of occupations or sectors.

As regards data protection and data privacy, profiling algorithms also carry the risk of being in conflict with the GDPR and the case law of the European Court of Human Rights and the Court of Justice of the European Union. Importantly, these legal bases provide no ready-made ‘checklist’ as to which data can be used, nor which algorithms can be implemented. Impact assessment of algorithmic profiling or job matching tools based on algorithms must therefore take place on a case-by-case basis that takes into account the impact of the algorithm on the citizens. Governments most often disregard the need for these impact analyses and entire profiling algorithm are therefore at risk of being shut down, such as in the Austrian case in 2020.

Impact assessments should first stress the necessity of using privacy-violating profiling algorithms. This can be justified in order to comply with a legal obligation to which the public authority is subject or for the performance of a task carried out in the public interest. The proportionality and fairness of profiling algorithms must also be checked and ensured. Proportionality relates to whether the ends justify the means.

For instance, collecting and analysing data carries a cost, in terms of privacy, which must be compensated by clear gains in accuracy. One should therefore not feed the algorithm with variables that have little explanatory power. Fairness concerns imply that one should ensure that profiling algorithms are not discriminatory. This is not straightforward. Profiling algorithms classify the unemployed based on the typical behaviour observed among other jobseekers with similar characteristics. As a result, individuals from social groups that are traditionally the least attached to the labour market will be profiled as high-risk individual more often than the rest.

While this behaviour of profiling algorithms seems intuitive, research has found that among jobseekers who happen to quickly find a job, those from foreign origin are more likely to be misclassified as high-risk individuals ex-ante than natives.

The fairness condition therefore seems hard to meet for profiling algorithms. Last, profiling algorithms should only use data that is up to date and relevant and, importantly, one should ensure that jobseekers and PES counsellors who use the algorithm have a good understanding of its functioning and limitations.

Whether or not the use of an algorithm is legal must be continually assessed before, during and after development and implementation. In a working paper based on deliverable 2.2 of the HECAT project, we therefore propose a model for designing algorithms to sum up these considerations. The model is circular in order to illustrate that the assessment should be continually updated.

A proposed model for designing algorithms

*Source: Working paper based on HECAT deliverable 2.2*

“Working with not on the unemployed”

Given these shortcomings of state of the art profiling tools, our European project HECAT puts the unemployed persons and their aspirations and needs centre-stage. It aims at building a sustainable digital platform “My Labour Market” which provides both information on the estimated length of time before one exits the unemployment record and a visualisation of labour market opportunities according to one’s job quality preferences. This digital platform, to be piloted at the Public Employment Services in Slovenia, builds on extensive sociological fieldwork on unemployed persons and case workers. This tool will not sort jobseekers into profiling groups associated with specific services and labour market measures. Instead, we believe that well-informed jobseekers will make the best choices for themselves.

[1] The dimensions are: pay and other rewards, intrinsic characteristics of work, terms of employment, health and safety, work-life balance, representation and voice, distance to work.

About the Authors

Clément Brébion, postdoctoral researcher, received his PhD in economics in November 2019 from the Paris School of Economics. His main research interests are labour economics, economics of education and industrial relations. He has a particular interest into comparative research. More recently, he started working on the EU H2020 project HECAT that aims at developing and piloting an ethical algorithm and platform for use by PES and jobseekers.

Janine Leschke, political scientist, is prof MSO in comparative labour market analysis. Her research interests comprise issues such atypical work, job quality, labour mobility and migration, youth unemployment, as well as gender. She is currently the Danish lead partner in the Horizon 2020 project HECAT, participant in EuSocialCit and one of the editors of Journal of European Social Policy.

Photo by Campaign Creators on Unsplash

March for Gender #2: The Gendered Impact of Covid-19

By Maha Rafi Atal

◦ 5 min read ◦

Most years, International Women’s Day is greeted by articles highlighting both progress made towards gender equality, and the distance still to close. 2021 is different. This year, organizations from the European Parliament to UN Women have instead drawn attention to how women have been pushed backwards – economically and politically – during the coronavirus. It has been “a disaster for feminism,”and a “great amplifier” which has exacerbated existing inequalities and unraveled tenuous gains. What does the research show?

First, the global economic contraction of the past year has disproportionately harmed women. In the United States alone, more than 2 million women have dropped out of the labor force altogether, a regression to 1988 participation levels, erasing a generation of gains.

Globally, women account for 54% of jobs lost during the pandemic, even though they make up only 39% of the global formal workforce.

Women bore the brunt of job losses in 17 of the 24 member-states of the OECD in 2020, and in South Africa, a survey found that two-thirds of workers laid off or furloughed in the first wave of the pandemic were women.

In part, this is a reflection of the sectors women work in, such as travel, tourism, restaurants, and food production, which have been largely shut down over the past year.

Women are also more likely to be employed on precarious or zero-hours contracts within these sectors, which made them vulnerable to job cuts, or in informal roles which left them outside the reach of government income-support schemes.

Finally, 190 million women work in global supply chains, including garments and food processing, and these industries have contracted as buyers either withdrew orders from suppliers during the recession, or sought to re-shore production closer to home. Labor market dynamics also mean women who stayed in work are among the most exposed to contracting the virus itself. A majority – estimates range from 67 to 76 percent – of the global health care workforce are women.

Yet only one quarter of the gendered discrepancy in job losses can be explained by the sectors where women are employed. Far more significant is the burden of care labor, both paid and unpaid, which disproportionately falls on women in both developed and developing countries.

Working mothers in the United Kingdom, for example, are 50% more likely than fathers to have either lost their jobs or quit in order to accommodate the responsibilities of caring for children with schools closed, with European women doing on average twice as much care labor as men during this period.

Over a million women in Japan left the job market in the first wave of the pandemic due to childcare needs at home, erasing tenuous progress the country had made towards workplace gender equality in the last decade. This unequal weight of the pandemic builds on pre-existing inequalities, as women are lower earners in many societies, meaning their jobs are considered a lower priority – by both employers and households – in times of crisis.

This economic crisis is not just a blow to women’s economic position, but to their political freedom. The “Local Diaries” podcast in India recounts the stories of women whose personal, political and sexual freedoms have evaporated as they have been locked down at home. As in pandemics past, covid-19 has seen a significant spike in domestic violence, femicide and other gender-bases violence in countries under lockdown. These include including developing countries like Nigeria, Argentina, Brazil, India, Pakistan, and China, and developed countries including the United States, the United Kingdom, France, Ireland, Lithuania, Sweden and Italy, a reminder that the home is not a safe place for many women. UN Women has referred to these spikes in violence as the “shadow pandemic.”

Moreover, despite early warnings from international organizations and women’s rights advocates, many countries shut down or diverted resources away from reproductive health care during the pandemic, leading to a rise in maternal deaths, unsafe abortions and pregnancy-related deaths. Finally, lockdowns themselves – and the expansion of policing and military powers associated with their enforcement – can themselves pose a risk to women, as police forces can themselves be significant perpetrators of violence against women, and as governments take advantage of these powers to suppress political organizing, including feminist organizing, as seen recently in both the UK and Poland.

At the same time, in a punishing political environment, women and feminist organizations have been at the forefront of pandemic response. The Chilean feminist movement has released a useful guide for governments and employers for responding to the pandemic in a gender-just way, while the Indian Kudumbashree women’s collective organized grassroots community kitchens and takeaway restaurants to provide food and employment to women, especially migrant women, during the country’s shut down, and repurposed textile micro-enterprises, largely women-owned, for the manufacture of PPE.

Despite calls from international experts for governments to respond directly to the crisis facing women by keeping services for reproductive health or shelters for victims of gender-based violence open, targeting cash transfers to women in informal employment and providing for paid child care, UNDP reports that only 12% of governments have adopted adequate gender-sensitive measures in their pandemic response.

Meanwhile, employers who have disproportionately laid off women in the crisis now report that gender equity will take a backseat to restoring their financial sustainability as the pandemic ends. This is made more difficult by the fact that some governments, such as the UK, have suspended requirements for companies to report on their gender pay gap or comply with other equality requirements, as part of pandemic support.

In our own research on corporate responses to covid-19, we found brands advertising luxury fashion goods to women and presenting the pandemic lockdowns as a welcome relief from labor in which women could enjoy them, a regressive image that shows how women’s work is still seen as frivolous and extraneous.

This International Women’s Day, then, we must reflect not on what progress we have made or can make, but on how women, internationally, can recover what we have lost.

About the Author

Maha Rafi Atal is a postdoctoral research fellow at the Copenhagen Business School, where her research focuses on corporate power, corporate social responsibility and corporate influence in the media. She is a co- Investigator on the Commodifying Compassion research project. http://www.maha-rafi-atal.com

Photo by Giacomo Ferroni on Unsplash

Distraction and manipulation: the two horsemen of the digital economy

By Jan Michael Bauer

◦ 2 min read ◦

Even before COVID, people have spent more and more time online. Particularly mobile devices have become a large part of our daily routines and for many there are few moments when the phone is not within direct reach. While studies have shown that even teenagers think they waste too much time online, surprisingly little is done to stop this trend.

But how did we get here? Several dovetailing factors enabled this development and give me little hope that this trend will slow down any time soon. While technological advancements in mobile internet and device components were necessary conditions that allow for an easy and enjoyable interaction with platforms and services at all times and places, the real champions of compulsive internet use are social and data scientists driven by monetary incentives and unrestrained by a lack of proper ethics training.

Despite the frequent regrets about the many hours wasted on the internet, people are struggling with self-regulation and apps, like “RescueTime”, with to sole purpose to block oneself from using other apps are becoming increasingly popular.

While internet addiction has not been officially recognized as a disorder by the WHO, close parallels can be drawn to officially acknowledge gaming and gambling addictions.

And this is certainly no coincidence as tech companies hire psychologists and designers to make their products and services as tempting as possible, frequently borrowing elements from the gambling industry. However, even though some tweaks based on the knowledge of capable social scientists will increase user engagement, much more can be learned about consumer behavior and how to manipulate it through the application of the scientific method itself. The use of experimentation, collection of big user data and application of machine learning algorithms are the big guns in the fight for user attention and their money.

All these efforts are used to make social media more “engaging” but ultimately sales and advertising campaigns more effective. To do so, user interfaces and features are explicitly designed to grab attention and contain what has been termed as “dark patterns”. Design elements that often tap into the subconscious decision-making processes and therefore manipulate user through purposefully curated interfaces. While such practices benefit the company, they can have detrimental effects on individuals and society as a whole.

We know that individual choices reflect individual preferences only under certain conditions, including the absence of deceptive choice architecture or marketing messages. Hence, I can’t stop wondering about the opportunity costs and side-effects of these miraculous little devices in our pockets that have grown into an ugly hybrid between a snake oil salesman and one-armed bandit.

We have free markets based on the belief that they create value for society and make people better off by efficiently satisfying their needs. The recent U.S. opioid scandal has shown that for some products, sellers’ profits might not be positively related to consumer value. It certainly gives me pause that the best offline equivalent to the “RescueTime” App is probably the Betty Ford Clinic.

We are faced with many pressing issues that would require our full attention, while people are increasingly plagued by credit card debt, the planet is suffering from overconsumption and we spent 30,000 years alone, watching Gangnam Style on YouTube.

Regarding the larger point that any efforts against these trends would hurt innovation, jobs and growth; let us take one step back and point out that the Western world has made it an imperative to ensure individual property rights and outlaw the use of violence with the explicit goal to increase investment and productivity. People can just do more good stuff, when they do not have to spend time protecting their property and family. Given our current technology and knowledge from the behavioral sciences, I think we have seen enough and should start treating distraction and manipulation as similar threads to human flourishing.

So, what could we do? In the short run, we need to find ways to reduce the stream of big data feeding these efforts, force these practices out in the open and raise awareness about their use and effects, and find effective regulation to limit manipulation efforts in a dynamic attention economy. In the long run, we probably need to go beyond those patches as these issues not only hurt individual lives and careers but also the fabric of our democracy.

About the Author

Jan Michael Bauer is Associate Professor at Copenhagen Business School and part of the Consumer & Behavioural Insights Group at CBS Sustainability. His research interests are in the fields of sustainability, consumer behavior and decision-making.

Source: photo by ROBIN WORRALL on Unsplash

Tag: #data

Sustainable labour market integration: challenges and advancements in algorithmic profiling of jobseekers

By Clément Brébion and Janine Leschke

A proposed model for designing algorithms

“Working with not on the unemployed”

Further readings

About the Authors