Publication: Investigating Hospital Service Robots: An Observational Study About Relieving Information Needs at the Hospital Reception
Authors: D. Sommer, S. Fischer, F. Wahl
Published in: International Conference on Human-Computer Interaction (HCII), pp. 395–404 (2024)
DOI: 10.1007/978-3-031-61932-8_45
In brief
A hospital reception is a bottleneck: questions from patients, visitors and external parties converge here, and staff answer them almost by the second. Service robots are meant to relieve exactly this point – but how much they actually take on in real operation has barely been documented.
This study closes that gap with field data. Over one week in September 2023, at the reception of a rural Bavarian hospital already equipped with service robots, we observed 1,703 interactions – and compared what staff handled with what the robot handled.
Why reception is a bottleneck in the first place
Before reading the numbers, it helps to look at the workplace itself. The reception is a building's first point of contact and at the same time one of its least predictable stations. Whoever walks in brings their request unfiltered: finding a ward, asking about visiting hours, looking for the right person to talk to, or simply getting their bearings in a confusing building. These requests do not arrive evenly but in waves, and they overlap with tasks that already demand attention – phone calls, handovers, administrative work.
For staff, this means constant switching. Each brief piece of information is trivial in isolation, yet together they tie up the focus and time that are then missing elsewhere. In a sector under chronic staff shortage in particular, every minute matters that can be shifted from routine questions to more demanding work. This is precisely where the idea comes in to hand part of the recurring requests to a service robot. The decisive question is not whether this is technically possible, but whether it provides noticeable relief in everyday operation – and that can only be answered in live operation, not in a lab.
The results
Of all interactions, reception staff handled the majority (89.9 %), while the service robots took on 10.1 %. The timing is telling: human staff were most in demand between 10:00 and 15:00, the robot rather between 13:00 and 19:00 – so it also covers fringe hours when fewer staff are available. Most interactions lasted under a minute and predominantly concerned orientation – for both the robot and the staff.
On balance, robot use saved 2.15 working hours during the study period. The setup used a Pepper (for gestures, touch interaction and FAQs) and a Temi, which Pepper could summon for autonomous guidance to points of interest.
What stands out is less the size of the share taken on than the kind of tasks it covered. The robot stepped in where requests are short, recurring and easy to structure – exactly the orientation questions that staff answer most often and most casually. The shift of the robot's activity into the later hours also shows that relief does not necessarily arise where the load is highest, but also where staffing resources grow scarcer. The division of labour thus played out not as competition but as complement: the robot took on a defined strip of the request spectrum, while staff remained responsible for everything else.
How to read this
10.1 % and 2.15 hours may sound modest at first – and that is exactly where the study's value lies: it is honest. It shows that today's service robots make a real but limited contribution at reception, and it clearly states that more capable robots and more comprehensive studies are needed to realise the full potential. That is solid research rather than marketing promises.
A one-week observation at a single site is deliberately not proof of a broad effect, but a snapshot under real conditions. Its contribution is to test a claim that is otherwise made all too sweepingly against measurable behaviour: a robot at reception is neither a mere gimmick nor an instant staffing solution. It is a tool with a clearly bounded scope of effect – and the value of a study like this lies in measuring that scope soberly rather than overstating it. Knowing today's limits lets you decide more deliberately where deployment pays off and where it does not.
Why we run studies like this
At Athegus we believe trust in robotics is built only on solid, self-critical evidence. This study shows where today's technology stands – and where the levers for the next generation lie: better robots, intelligent task allocation and a platform that makes humans and machines work together sensibly, rather than putting up a robot for its own sake. That orchestrating layer is our product.
The interplay of two devices in this study – a Pepper that receives requests and a Temi summoned for guidance – is a foretaste of the real problem at scale: as soon as several robots, systems and people work together, it is not the individual device that determines the benefit, but the coordination between them. Tasks have to be sensibly allocated, handovers cleanly executed and responsibilities kept clear. This orchestration is not an accessory but the point at which individual robots turn into a reliable contribution to operations – and that is exactly where we come in.