Page 268 - Artificial Intelligence for the Internet of Everything
P. 268

246   Artificial Intelligence for the Internet of Everything


          13.2.4 Joint Cognition
          Humans are nearly always components of complex systems, and we are
          obviously unique in our capabilities. Although we are sometimes passive
          subjects in a system, more often we play active roles ranging from sensing
          to actuation to control. Often, systems are most frustrating when they force
          us into passive roles rather than providing us with agency. The problem of
          automation has been shown to be lack of understanding of the system by the
          operators leading to errors pointing to problems in the design of such joint
          cognitive systems (Bainbridge, 1983; Norman, 1990). When we wish to
          design systems in which human actors are components, our representations
          must go beyond artifacts, to include models of human behavior. When
          humans (or other autonomous agents) participate in the control loop for
          a complex system, it is essential that we know which information should
          be suppressed, what should be shared, and how that information should
          be presented in context. Compared to machines, humans are slow and error
          prone, but without our flexibility and global understanding systems become
          brittle and liable to fail (Woods & Hollnagel, 2006).



          13.3 SYSTEM DESIGN IS A RECURSIVE PROCESS

          In this section we illustrate the well-known claim that engineering and
          design are fundamentally recursive processes. Although it is more obvious
          in systems of systems like the IoT, the claim is not specific to systems engi-
          neering. Rather, as was noted earlier, recursive features like problem
          decomposition have always been present in engineering, but for small pro-
          jects they could be managed informally with human expertise. Regardless, it
          is the complexity of modern hybrid systems that forces us to acknowledge
          and understand this aspect of engineering and design theory.
             An example helps to fix ideas. Suppose Alice is tasked with designing a
          heating, ventilation and air conditioning (HVAC) system satisfying various
          requirements (maintains set temperature  1°C, cooling rate of 2°C/hour,
          etc.), the task and its requirements might come directly from the customer,
          or internally from another part of the design process.
             Alice decomposes her system into four components: a heater, AC, ther-
          mometer, and a controller (Fig. 13.1). She designs the controller herself and
          locates an off-the-shelf thermometer that will meet her needs. Then (the
          recursive step!) she tasks Bob and Charlie with designing the heater and
          AC, respectively. Some of Alice’s requirements can be passed directly to
   263   264   265   266   267   268   269   270   271   272   273