Establishing trust is not just a process of how onshore control center should believe the good the automation ships can perform at sea. In his paper <The ‘problem’ of automation: Inappropriate feedback and interaction, not ‘overautimation’>, Norman advises that automation is at an intermediate level of intelligence. It is powerful enough to take over control that which used to be done by people, but not powerful enough to handle all abnormalities [61]. To offer a powerful automation, providing appropriate feedback that occurs among actors is essential. For example, in Figure 3 automation ships could maneuver themselves at sea. However, such auto-maneuvering should provide feedback to the onshore control center as well as enable people in the onshore control center to monitor important information regarding communicate with ship owners and vessel traffic services. Also, automation ships also need to be monitored by the vessel traffic services due to the busy traffic at sea, coastline, and harbor. In this manner, both onshore control center and vessel traffic services may share the same data regarding the automation ships and offer a chance for ship owners to make decisions.
From an ANT point of view, all actors now speak out their concerns regarding the automation. Both vessel traffic services and onshore control center need to monitor the automation ships. Automation ships need to maneuver regarding the planned tasks. Ships owners concern the efficacy of the planned tasks as well as the status of automation ships. However, the actors-in-the-loop has the same goal – the safety of the vessel. Trust is established and configured through the interactions among different actors through several means – monitoring, communication, and feedback. Thus, this dynamic establishing and configuring process enable automation, and human (navigator) shift their roles in the control of automation ships by allowing onshore control center to control the vessel if necessary. This can avoid onshore control center or vessel traffic services lose awareness during the monitoring behavior and mistakenly trust the automation. In such manner, it is important to introduce how to implement the actor-in-the loop when designing an IoT platform.

Implementation of actor-in-the-loop

ANT enables each actor to speak out interest and adjust such an interest to meet the other interests of other actors in the same loop to make the IoT platform useful. As a multidisciplinary field, IoT development is more than pure engineering work. It requests a systematic way to enable academic outcomes, such as methodology, theory, and method from the other field also can contribute to the development process. Thus, the implementation of actor-in-the-loop, therefore, should focus on three important points.
It is important to identify firstly who is an actor in the IoT platform. In this maritime project, the actors as discussed are marine professionals, IoT platforms (including Vessel traffic services, satellite, onshore control center), automation ships, and ship owners. All these actors together established the actors-in-the-loop for enabling the remote control of automation ships successful.
Second, it is important to investigate the interest of each actor. This requests trained field workers to engage with different actors to observe, interview, and note the important issues for the design process of IoT platform. For example, Sharp et al. [20] present the importance of fieldwork that can help developers to identify central requirements for development and control the quality of the software systems. In line with this, we assert that marine professionals can elaborate the most important information for their work practices. With that information, it is not difficult for developers to decide how to choose relevant technology to display information for relevant actors. Further, it is not difficult to decide which sensor should be used. Thus, it is possible to design the onshore control center to remote control the automation ships as a useful and important part of the IoT platform for the automation purpose in the maritime domain.
Third, it is important to investigate the dynamics of actors-in-the-loop. Trust is established through the adjustment of different interests of actors in the loop. However, to bear in mind that each actor has its interest. Thus, it is vital to show up when, where, and how different actors establish trust in the loop of actors to arrange most suitable sensors, technologies, and devices to support such trust in the IoT platform. In this case, the IoT platform becomes not only a collection of hardware and software tools but also has a sense-making process of collection those tools for a useful purpose. In this manner, the development process of IoT platform can be sustainable. That is we might less waste hardware and software in the development process if the people who use the IoT platform refuse or dislike the results which the HITL model provided – or we call it maritime work through an imagination by IoT developers.
To summarize, the development of IoT is a process to develop an information space for the different field to contribute together [62], [63]. This requests several different roles be involved, such as ethnographer, systems developers, project manager, and project owners and so on. Thus, the development process of IoT is a process of sense-making of the sociotechnical solution to support technical work of end-users in their working context. Therefore, it is needed to elaborate how to design artifacts in the IoT platform.

Discussion

Subject, object, and interactive relations in IoT development

A successful IoT development requires a focus on material and nonhuman actors as constituent elements of the social and technical, and in which there is a shift from a focus on ‘epistemology’ (crudely, how is it we can know something) to ‘ontology’ (crudely, what things can exist, or in ANT terms, how are things enacted?). On this score, ANT has been integral to the so-called ‘material turn’ in technology studies.
In tradition, development requirements for IoT platform are based on the work-as-imaged (what we think people do) through developers and requirement specialists. ANT gives an alternative to force us to think about work-as-done (what people do) and collect these pieces of work practices to shape the IoT platform towards as a useful and helpful tool.
If material and nonhuman are artifacts (of course they are), then the social meaning and work practices around those artifacts make the artifact computable and designable [47]. In this matter, the design of IoT platform becomes to investigate how work practices around an artifact connect to other artifacts and make sense of it. Thus, actors can always play in the loop to ensure the social IoT is meaning for the people who use it. It importantly improves the process to use human as subjects for testing and validating purpose [64] but engages the humans who can give voices to the product they are designed for [65].  Moreover, the theory of ANT provides the very idea of the social in IoT has to be interrogated as a construction rather than assumed as an explanan, doing all the explaining in requirement specification in IoT design and implementation.

The combination of social and technical work