Requirements and recommendations for human-centred human-AI collaboration design in manufacturing domain

In the factories of the modern world, the rhythm of production is changing. For decades, the goal of industrial innovation was to make machines faster, cheaper, and more automated, often treating human workers as components to be replaced by robots. This era, known as Industry 4.0, brought incredible efficiency but also raised concerns about the well-being of the people on the floor. A new vision, called Industry 5.0, is now emerging to correct this balance. It proposes that technology should not just replace humans, but work alongside them, enhancing their capabilities while protecting their safety and dignity. At the heart of this shift is the concept of human-centred artificial intelligence. This approach does not view AI as a black box that simply executes commands, but as a partner that must be designed to understand human needs, values, and limitations. The challenge for engineers and designers is no longer just building a smart machine; it is figuring out how to build a relationship between a person and a machine that feels natural, safe, and productive.

A team of researchers from Tampere University in Finland set out to solve this specific puzzle. They wanted to move beyond general rules for how humans and computers interact and create a practical guide specifically for the messy, noisy, and high-stakes environment of manufacturing. To do this, they did not rely on a single experiment or a computer simulation. Instead, they gathered a wide range of existing knowledge and combined it with fresh insights from nine experts who work with robotics, software, and AI design. These experts, all based in Finland, shared their experiences through detailed interviews, discussing what works, what fails, and what is missing in current factory designs. The researchers then wove these interviews together with a review of academic papers and industrial reports, looking for patterns that could turn into a clear set of instructions for designers.

The result of this work is a new framework that shifts the focus away from the AI system itself and toward the collaboration between the human and the machine. The researchers found that designing a successful partnership requires looking at eight distinct areas, or dimensions, rather than just the software code. First, before any design begins, teams must clearly define the goal of the collaboration and understand the specific factory environment, including factors like noise levels and dust, which can change how a machine is used. They must also assess whether the company and its workers are ready for this change and ensure that ethical concerns, such as worker privacy and the risk of bias, are addressed from the very start.

Once the groundwork is laid, the design must focus on how the work is actually divided. The study suggests that tasks should be allocated based on what each partner does best. Machines are excellent at repetitive, dangerous, or data-heavy jobs, while humans bring creativity, flexibility, and the ability to handle unexpected situations. The goal is to let the machine handle the burdensome work so the human can remain an active, engaged decision-maker, rather than a passive monitor who simply watches a screen. This leads to the second major finding: the human must stay in the loop. If a worker is reduced to merely watching a robot work, they may lose their skills and feel disconnected from their job. The design must ensure that the human remains the leader of the team, with the AI acting as a supportive teammate that augments human ability rather than replacing it.

The researchers also identified that the "team" itself needs to be built carefully. This involves matching the right human skills with the right machine capabilities and ensuring that both parties have the training to work together effectively. A critical part of this is the machine's ability to explain itself. The study emphasizes that AI should not just give an answer; it must be able to communicate why it made a decision, what its limitations are, and when it is unsure. This transparency is vital for building trust. If a worker does not understand why a machine is acting a certain way, they cannot safely intervene or rely on it. The communication between the two must be a two-way street, where the human can ask for clarification and the machine can proactively share important information before a problem occurs.

Beyond the immediate interaction, the study highlights the importance of the physical and digital infrastructure that supports the team. This includes the sensors, displays, and data networks that connect the worker, the machine, and the factory floor. This infrastructure must be designed to provide clear, timely information without overwhelming the worker with too much data. Finally, the researchers stress that the organization itself has a responsibility. Companies must provide continuous training to help workers adapt to new technologies and must establish clear rules about who is accountable for the work. The study concludes that successful human-AI collaboration in manufacturing is not a technical problem to be solved once and for all, but a continuous process of adaptation. It requires a socio-technical approach that considers the human, the machine, the task, and the organization as a single, interconnected system. By following these guidelines, manufacturers can create environments where technology supports human well-being and resilience, ensuring that the future of work is not just smarter, but also more humane.