Fiction? Or soon reality?

In a digital administration, artificial intelligence is meant to take work off human staff and speed up processes. That also means: filtering, sorting, prioritising applications — or even deciding on them. But how much power is too much?

In February 2026, the study

“Digital decision-making models in e-government systems”

was published by Dr. A. Rahman, Prof. L. Steinberg and Dr. M. Kovács.

The central question: how can algorithmic decision-support systems be embedded in administrations so that they stay compliant with the rule of law, transparent, and accountable?

The study draws a clear distinction between decision support, partly automated decisions, and fully automated administrative acts.

The more automated the algorithms’ decisions become, the higher the requirements for transparency, documentation, the ability to object, and human override become. Because only when all four are guaranteed do AI decisions actually meet the standards of the rule of law.

One problem stood out especially clearly in the study: diffusion of responsibility.

Rahman, Steinberg and Kovács show: once algorithmic systems are introduced, it becomes unclear who’s professionally and technically responsible, who’s liable for wrong decisions, and how objections get handled.

But how could it be done better? The researchers propose a three-tier model:

1. Every algorithmic component needs a named person responsible for it professionally.

2. Only decision logics that are comprehensible are used in sensitive administrative procedures.

3. Humans make the final decision in risk-laden cases.

What do you think about algorithms making decisions: opportunity, or absolute risk?

Source: Rahman, A.; Steinberg, L.; Kovács, M. (2026): Digital decision-making models in e-government systems, published February 2026 (ResearchGate).

Image: This image was created with AI.

You can find this post on LinkedIn here: https://www.linkedin.com/feed/update/urn%3Ali%3Ashare%3A7430619356121886720