A recent study based on expert interviews takes a look under the bonnet of the digital welfare state by asking what kinds of motives and strategies guide the design of digital self-services. The study examines the design of online public employment services and the forms of structural injustice these services may create for certain groups of people.
“We wanted to understand how public services’ preparation for the use of artificial intelligence (AI) is shaping public administration and online services. Previous research suggested that AI is still used only to a limited extent in public services, both in Finland and elsewhere,” says project leader and senior research fellow Jaana Parviainen.
Public services are being developed with data as the priority
The researchers interviewed a wide range of designers, IT company project managers, civil servants, programming and data analytics professionals, and decision-makers involved in digital development, including politicians. The interviews revealed that digital platforms are primarily viewed as new administrative tools. Their purpose is to reduce the need for costly face-to-face services, collect data to improve administrative efficiency, and increase public administration’s access to information for information management purposes.
“The expert interviews highlighted the administration’s growing need for data-driven governance and the collection of user data through digital platforms. The aim appears to be to create ‘data infrastructures’ or ‘data pools’ that enable the state to develop algorithmic governance and data-driven decision-making,” Parviainen notes.
One of the study’s key findings was that digital self-services are not the result of a technological breakthrough or an ideal solution to job seekers’ needs. Rather, they represent a compromise shaped by the political, administrative, economic, legislative, and technical factors underlying the design process.
Digitalisation can increase structural inequality
The researchers were interested in the kinds of epistemic benefits and harms that data-based online services may create from the perspective of citizens and service users.
“First, job seekers are almost compelled to use digital self-services. Experts recognised that for some groups, such as certain migrants, using these services without assistance from other people is almost impossible. The absence of online banking credentials may prevent access to services altogether, and users may not always recognise the information security risks associated with sharing credentials,” says postdoctoral researcher Sini Teräsahde.
At the core of employment services is an attempt to solve the problem of labour market mismatch by helping the right employees and employers find each other.
“Jobs are recommended to individuals based on their skills profile, but the available tick-box categories do not necessarily allow people to describe their actual competences. For users, the algorithmic recommendation system functions as a kind of ‘black box’,” Teräsahde explains the findings.
The black box of algorithmic systems
A black box means that users of digital employment services cannot know how the algorithmic recommendation system operates. Neither experts nor users necessarily know what kind of data are used to train the algorithms behind the system. As a result, job seekers may find it difficult to understand why certain jobs are recommended to them. The interviews revealed that even some developers struggled to explain the operation of the recommendation algorithms at a basic level.
“We concluded that users’ lack of interpretive resources regarding how the system functions may reinforce inequalities between job seekers. Digital platforms appear to be designed primarily for digitally competent citizens who can access services anytime and anywhere through mobile devices,” Teräsahde summarises, and continues: “These users are also better positioned to learn how the systems operate and can play with them to their advantage.”
So far, the structural effects of algorithmic systems on epistemic justice have been studied relatively little in Finland and internationally as well.
“Unfortunately, our findings suggest that improving citizens’ digital skills alone is not a solution to the problems created by digitalisation. If we want to address issues of inequality, we must also tackle the structural challenges produced by digitalisation in society,” Parviainen concludes.
Original peer-reviewed scientific article:
Parviainen, J., Teräsahde, S., Alanen, P. Koski A. & Rantala, J. (2026) ‘This is a matter of life and death for efficiency’: the implications of datafication within public services for epistemic (in)justice. AI&Society https://doi.org/10.1007/s00146-026-03157-9
Further information: jaana.parviainen@tuni.fi & sini.terasahde@tuni.fi