Human-Machine Collaboration in Decentralized Physical Infrastructure (DePIN)
Human Factors in DePIN
In decentralized physical infrastructure (DePIN), a decentralized network of edge devices cooperates to provide a service that conventionally is delivered by a centralized system. Advantages include higher efficiency resulting from shared processing and data, better service by local adaptation, and innovation through an open platform, control, and analysis. Human-machine collaboration is common in many sectors and situations in the operation, control, and management of physical infrastructure that provides such services. However, in many examples, we have observed humans becoming passive or being excluded when machine performance exceeds human performance or domain knowledge is required. In DePIN, machines can collaborate with and empower humans for good results. However, the spatial and temporal fragmentation of DePIN is also a source of challenges to the participation of the various humans in human activities that create, test, dispatch, and manage control of DePIN.
Human activity overall is oscillatory, as opposed to continuous. Motivation, diet, time management, and learning lend themselves to a hysteresis model. Continuous operation and exposure to open-loop or arbitrary noise violate fundamental human values and operate against the interests of the organization and historic expectations. The necessary physical lattice of DePIN imposes an opportunity cost that can only be justified by a clear efficiency dividend. The lattice of DePIN limits natural interaction with entities defined, or not at all, by the lattice. Similarly, knowledge and understanding of not only the items cared for by DePIN as a whole but also their relationships are moderated. Consider, for example, the work which constructively shaped both the cartoonist and the lamp post over a century; the loss of the latter without the other would be a profound and irreversible change, at least to them. We distinguish three main areas of human factors. First, humans generally desire local personal feedback. Secondly, we consider specifically tasks that have specific physical, human, organizational, or societal requirements or constraints beyond merely the hardware functional and performance requirements. These can be positive tasks (construct, renovate, dismantle, interpret, contemplate, appreciate, maintain) or negative tasks (judge, impede, resist, sabotage). Thirdly, we consider types of people whose praxis benefits DePIN. These include the user market, providers, and service providers. Service provision includes habilitators, maintainers, inspectors, regulators, and laypersons performing lay tasks. Service providers require training, information, and decision support, including expert systems.
Shared processing of data and sharing of data are distinguishing characteristics of DePIN. Data is drawn from many distributed sources, not only other entities in the collective or lattice but also from individuals not otherwise affiliated, including many people. People need to be engaged in the goals of the collective. On a personal level, connected operation also provides human benefits for mentoring, development, and service. Unlike many internet interactions, the two-way street of information flow between a person and the collective ensures that there is also a corresponding benefit to the DePIN for long-term knowledge and intuitional updates. User, provider, and the collective are all serviced. Superfluous or incorrect data from scammers or from outdated sources requires particular attention to the principles of fake news. Legitimate data sources that expect a return need to be attracted into the collective. Decency and privacy require ordered and authorized data traffic. Economically, incoming data has a cost. For either hardware or persons, utilizing the local resources that drive DePIN is cheaper than including the chance of a journey to a random location. User-led action needs to be paid in proportion to either the human or financial cost incurred by the collective. Collectively, DePIN necessarily operates against the opposing widely adopted business model.
Cognitive Load and Decision-Making
In recent years, the relationship between cognitive load and decision-making has received much attention in the HCI literature. Cognitive load is the “mental effort needed for a task” and decision-making is “the cognitive process of selecting a course of action from among multiple options in order to solve a specific problem.” In discussing universally accessible design, Mayhorn et al. maintain that “individuals can be overwhelmed by the amount of information presented to them or the complexity of the cognitive tasks needed to use interfaces”, influencing decision-making in an interface. In his divided attention model, Gray found that a high workload and insufficient display can contribute to increased workload, which, in turn, impacts operator performance. In task decomposition with accountable agents, cognitive load distribution was discussed and it was shown how complexity of simulation models contributes to large cognitive load in emergency management.
Researchers have developed models to describe how people are influenced by cognitive load when making judgments. Wright performed an experiment to explore the role of juxtapositional effects when lighting on photographs was varied and analyzed how different levels of cognitive load affected the judgments made by participants. They reported that an increase in cognitive load significantly influenced participants’ judgments, as well as participants’ ability to provide judgments regarding the stimuli. In another instance, Kruglanski and Mayseless presented participants with cognitive load conditions before measuring participants’ sense of independence, self-esteem, and influence of external forces. Their two experiments showed that, compared to the low-load condition, the high-cognitive-load condition was associated with lowered independence. These studies show that high cognitive load impairs human decision-making capabilities in judgment tasks that are unrelated to any immediate material gain.
Research in pervasive computing often focuses on creating smart sensors or smart systems on the infrastructure that supply valuable information for human operators in centralized intelligence support tasks. However, there is still a lot of untapped potential for intelligent infrastructure to perform actual decision-making tasks in decentral intelligence support. Going from central to decentral decision-making enables the quick and reliable provision of intelligent support for multi-agent systems from one single entity and hence can increase the robustness and reliability of the overall system while decreasing the cognitive load of human operators.
We introduce the concept of Decentralized Physical Infrastructure that Makes Decisions and coordinates the efforts of multiple human agents by selecting them for additional task-oriented ad hoc communication units. The focus of this work is on investigating how the cognitive load of human operators and the system can be adjusted to the complexities and criticalities of the task.
We introduce the research area of cognitive load management in Human-Machine Collaboration (HMC), where physical infrastructure is making decisions and coordinating multiple human agents, who are distributed in the area of task execution, by selecting them for associated ad hoc communication units. The focus of this work is to investigate how the cognitive load of human operators and the system can be adapted to the complexities and criticalities of the task. The system design is centered around principles developed for organizational processes, further developing them and demonstrating them in the novel domain provided by DePIN systems.
Definition and Types of Cognitive Load
Cognitive Load (CL) refers to the amount and capability of human cognitive resources utilized to solve complex problems. While cognitive resources required by machine components can be considered as computational resources, cognitive resources required by human operators are more context-dependent. Cognitive architectures such as SOAR, ACT-R, and BDI have been designed based on psychological theories of human cognition. Although such cognitive architectures are designed to help develop intelligent agents, we can use these architectures to develop models for estimating cognitive loads of human operators in various complex tasks. This can help a collaboratively interacting intelligent system to be more attentive and more effective in taking actions. It is essential to understand the definition of the term “cognitive load.” Cognitive Load (CL) refers to the amount and capability of human cognitive resources utilized to solve complex problems. It comes from information processing theory and is mainly used in the area of experimental psychology and human-computer interaction for designing educational materials and software systems to reduce cognitive loads of users performing given tasks.
Based on the types of cognitive load, there appear to be different ways to measure CL. Broad and complex human activities draw resources from perception, memory, reasoning, decision making, attention, learning, knowledge, and other cognitive functions. For example, researchers analyze the behaviors of problem solving, learning, and sequences of activities employing thought into these cognitive functions. Although information processing itself is performed at the level of detailed states and transitions, we can understand these states and transitions using higher-level categories. We can also categorize them as types of cognitive load such as intrinsic load, extraneous load, and germane load. Information processing consumes a variety of cognitive resources depending on the information types, mental processes, mental states, and task contexts. We analyze them through these categories.
Enhancing Efficiency through Human-Machine Collaboration
Real-time Monitoring and Predictive Maintenance
Real-time Monitoring
The era of the Industrial Internet of Things (IIoT) has introduced the possibility of gathering data from a diverse set of devices and exploiting such data to make fact-based judgments on the state of processes that used to be maintained through pure experience and running procedures. A component of IIoT is the capacity for smart machines to be monitored in real-time, with data collected and arranged for analysis that permits an understanding of machine operation under normal conditions, routine variations, and compelling deviations. This facility has received extensive attention in manufacturing and other process industries where continuous machinery operation is crucial for business, with studies on the collection and use of sensor data for monitoring machine health, whilst it became apparent that machine failures can be associated with, or preceded by, abnormal behavior driven by a degradation process.
Present human-machine collaborative systems have various limitations, such as the lack of consideration for human mental state, cumbersome learning of the human side, and difficulties in preventing human errors. By introducing artificial intelligence that can consider human mental processing, it is expected that task allocation will shift more to the field chief therapist. The actual collaborative tasks that are performed along with field chief therapists should be converted from physical assistance work to higher-level cognitive tasks, such as consultation, explanation, and relaxation. Collaborative work has effects, and positive reinforcement of human abilities occurs. This research aims to develop a system that enhances the suppression effect due to human participation and aims to help improve the operation of collaborative work. In this research, to achieve this aim, real-time monitoring that evaluates human operation for modifying the human-machine collaboration strategies and the strategies themselves was introduced into the developed system.
As a first step toward a theoretical discovery of the specification for such a system, we emphasized that creating two types of systems that collaborate in the process can prevent some negative effects of actual collaboration. This paper is the technical extension of the conference paper. Collaborative support is a concept that is quite close to collaboration in the relationship between human-robot interaction, and in implementing collaborative support, human participation is essential. Then, the approach must be either the passive participation of the person to whom the human is assigned or the strategy carried out by the image of negative feelings. The basic physical steps of the robot collaborator’s work must not change if the person carrying out the task is working on the problem, but may change under other conditions. The schematic model for the robotic intervention for human-robot interaction in two human affairs is reported. Robots should not pose a threat to the host, while anthropomorphic robots, while participating in the task, should try to promote the effect and preserve the actor’s role.
Real-time monitoring of manufacturing equipment for end-of-life predicting proposes many advantages, including increased safety, minimized downtime (hence increased productivity), improved process design through actionable equipment degradation data, reduced maintenance costs due to optimized planning and avoiding excessive preventative actions, and material usage and energy and emission savings. The integration of sensing technology and cloud computing allows for novel dedicated services for interconnected machines at unmanned sites where frequent physical checks and simple operations are typically carried out by humans. The availability of machinery monitoring data in the cloud for Industry 4.0 applications are major market drivers. As customers expect services related to the mobility of technicians, including the real-time tracking and dispatching of service cars to verify if necessary interventions are due, they are entitled to evaluate if investment in machine sensing and connectivity technologies are valuable.
Predictive Maintenance
The rise of Industry 4.0 has brought about technological transformation in the way that organizations conduct business. Through the leveraging of data through analytics and artificial intelligence, together with the unpacking of digital solutions, organizations are on the path of improving efficiencies and driving down costs. Making use of artificial intelligence being employed, organizations can pre-identify when a machine will have issues. In doing so, organizations avert incidents due to machine malfunction and mitigate the financial impact that follows afterward. Predictive maintenance would ensure that organizations could identify when a machine would need maintenance and tend to it ahead of its due period.
The concept of predictive maintenance works by leveraging the virtues of Industry 4.0 digital solutions to identify when a machine would be due for a maintenance check. Industry 4.0 digital solutions attract the spotlight for exploiting the technology of the Industrial Internet of Things (IIoT), which is the industrial application of the Internet of Things (IoT). IIoT constitutes components such as industrial sensors that can attach to industrial equipment, which generate large amounts of data that can be processed by AI. Cut to the chase, AI algorithms would analyze data sent back by industrial sensors attached to operating machines. Once the AI algorithms process the data, they would supply an indication of the status of the machine, thereby ascertaining if the need for preventive action would be necessary. If a machine would require preventive action, the AI algorithm would signal to the in-house technological team indicating the need for a check.
Sophisticated monitoring and control predates the modern age. What is unique about this modern age is the increased role of machines (including computers, of course) in all applications, small and large, simple and complex. In this paper and in related publications, we tout the idea of shifting responsibilities and tasks from traditional human referees and supervisors to new entities consisting of human-machine collaborations. Proponents present this notion in venues other than the academic and scientific literatures of their disciplines in a public-relations campaign to garner acceptance by potential clients.
Similarly, the broad notion of collaborating partners also predates the present. Such relationships are deep-seated in the social sciences, the arts, and the humanities where results more often than not rely upon joint efforts. Within the boundaries of the broad field of engineering, progress in a number of areas hinges upon successful partnerships with attendant beneficial synergistic efficiencies between humans on the one hand and machines running advanced applications software on the other hand. It is precisely this encouraging notion of such a human-machine partnership that we seek to communicate to the people and organizations that could benefit from advances in the real-time monitoring and control that we nourish in myriad ways.
