With the constant penetration of automation and robotics in industrial contexts, the nature of human tasks and involvement with technology is changing. The increasing intelligence and sophistication of systems enables human operators to not only manually operate them ("in-theloop"), but also to transition into a supervisory role ("on-the-loop"), where machines are monitored sporadically and over a distance. Various models within and across application areas to categorize the degree of automation and human involvement therein have been proposed. This paper investigates “the human in the loop”, by revisiting previous considerations of this term and by placing it into the context of human automation interaction in the production sector. Examples of current research projects are provided, in order to critically reflect on the relevance of the “Human-in-the-Loop” concept in the innovation of automated production processes.
Automated systems, even with the highest levels of automation, require some form of human involvement. Especially with high automation, the form of involvement has often been called “human in the loop” (HITL, or HIL). Given the rapid innovations of intelligent and automated systems, HIL has become a ubiquitous term for virtually all application fields. The common denominator of the term "Human in the loop" is that it refers to the involvement of a human to provide oversight, feedback, or intervention as needed in a process or system. The involvement of a human can be within decision-making or problem-solving, or the validation or correction of an automated system's output.
Recently, the notion of human centricity has been
nurtured by the Industry 5.0 initiative of the
European Commission [
To achieve this collaboration, human operators
are understood as highly relevant actors, who
should be enabled to engage in the
decisionmaking process of automated systems and
contribute their expertise and insights. In fact, the
human centricity concept of Industry 5.0 calls for
further research on the process of designing and
adapting intelligent technologies in work
processes with a focus on the human in the loop,
by empowering human operators to improve
productivity as well as well-being of workers
[
Given the importance of placing the human in the loop especially in this field of Industry 5.0, it appears relevant to come up with clear definitions of the human in the loop in human machine collaboration, to leverage the strengths of both agents. Figure 1 shows typical scenarios of Industry 5.0 where novel forms of user interactions within the human loop are currently investigated.
As outlined before, the term has been used in
many professional and practical fields, and as a
consequence, it has remarkably varying meanings
and connotations across domains. Even more
notably, HIL has often been dropped as a term, but
almost never accompanied by a concise definition
or terminological reference. Especially, the field
of industry 4.0 and 5.0 has only been roughly
considered with regard to human-in-the-loop [
In this paper, we make an attempt to come towards a more comprehensive understanding of the human in the loop for Industry 5.0, by reviewing the concept and its uses across research fields. We start out by providing an overview over the “etymology” of the term and its provenience. Then, we elaborate on some of the key aspects of the human in the loop, highlighting especially the multifaceted nature and different requirements and manifestations in the domain of Industry 5.0. We conclude with a critical discussion about the value of the “Human in the Loop” concept for different purposes.
Human factors engineering: The presumably
earliest appearance of the term “human in the
loop” has been introduced in human factors and
control engineering theory, where control loops
are used to describe the flow of decisions and
human-machine interactions ([
Simulation: Also in the field of simulation for
training and research purposes, the term is central
to describe simulations that are not only driven by
deterministic computer models, but that also offer
the involvement of humans to represent the
“unpredictable behaviour” of humans and to
evaluate the fit for purpose of systems and models
in real-world situations ([
Data analytics: Human-in-the-loop concepts
are also discussed in data analytics, where humans
are involved when “data problems” occur.
According to Doan [
Machine learning: Current references to HIL
mostly are related to the Human as an important
part to improve machine learning models [
Cyber-physical systems: Another stream of
research investigating human-in-the-loop
concepts relates to Cyber-physical systems, of
which Industry 5.0 will be a key application field
[
In the following, relevant aspects for appropriating and reflecting the human-in-the loop concept for human-centric design of Industry 5.0 are presented.
Depending on which of the above-discussed research fields is concerned, the identity of the human involved in the loop can be either direct users (e.g. in case of CPS usage) or domain experts (in case of machine learning and data analysis).
Especially Industry 5.0 processes can actually
involve different types of users at once. In the case
of the automated material handling vehicles
scenario depicted in Figure 1, the users to be kept
in the loop could involve a foreworker, a fleet
coordinator, an onsite worker, or a driver of a
semi-automated vehicle. In other cases, such as
the visual inspection of die casting or the overall
monitoring of production processes for quick
manual interventions in case of failure, the human
to be kept in the loop is mostly the worker in
charge of the quality control (cf. [
Models to create taxonomies for characterizing
the human (in the loop) may incorporate these
roles, along with other relevant accounts, such as
Cimini et al. [
It is not always clear, how metaphorical “the
loop” is thought to be. Depending on the research
field, a loop can be an implemented and
operationalized real-time system control loop [
For the human factors study on real-time
interaction with CPS, it is important to
differentiate between different relations of the
human in relation to the loop. Merat et al. [
Out-of-the-loop situations, where an operator
is moved out of a control loop due to automated
control, can lead to limited awareness of system
states [
Industry 5.0 scenarios are often characterized
by distributed teams of workers. This more
systemic view has been investigated early on in
human factors research. For example, Moray and
Hancock [
Cimini et al. [
There is an overall consensus that in virtually all situations and setups of future production environments, the human should remain at a central position. In this context of arguing for the significant necessity of human involvement, the term “human in the loop” has been convincingly used, while often in a rather superficial manner. Likewise, the “human in the loop” term as a key system characteristic highlighting human involvement for improving machine learning, data analysis or training outcomes.
As shown above, the human in the loop is a concept that has been used in various application contexts but less often been defined in a concise way. Despite a considerable number of papers carrying the human in the loop in their title, so far no comprehensive overview about the various uses of the term as such has been provided. This paper provides an attempt to review the term “human in the loop” from such a broad perspective and to gather relevant inspirations and remaining gaps for the application of this concept in the field of HCI in the Industry 5.0 domain.
An overall observation is that, although the human in the loop should be a genuinely humancentered aspect, the term has often been used by engineering domains. When considering the human in the loop as a way to communicate requirements and design human-automation interaction solutions in a project team, one can see the current gap that they mostly remain at a relatively generic level when it comes to the actual user interaction to be described.
We believe that in many Industry 5.0 use cases, an elaboration of the human in the loop concept would be beneficial as a conceptual design tool. This certainly applies for those types of scenarios that involve a rather small number of loops (e.g. quality control interfaces, see Figure 1) and where human interventions come sporadically (e.g. automated vehicle fleet management in logistics settings or human interventions in a highly automated manufacturing process, see Figure 1).
For other, more complex scenarios that can less easily be explained by loops and human placements therein, such as human-robot teaming and joint cognitive systems, the usefulness of a further elaboration of the term may be more questionable. Further research is encouraged to clarify the added benefits of the human-in-theloop concept for Industry 5.0 and beyond.
This work has in part been funded by AIT Austrian Institute of Technology, in the framework of the Lighthouse project “AI-enabled and sustainable Automation.”