Situated Intelligence

IF WE ARE to establish a new Science Of IoT, then it is vitally important to be certain that we are founding—or building—upon a firm substructure. In normal scientific language, the foundation, or basis of any argument is named as the hypothesis, and is often a subject’s supposition, or primary subject-matter.

In this respect, and in order to provide a focus for detailed analysis of IoT subject matter; we hereby provide a comprehensive definition for: Situated Intelligence. This new concept of Situated Intelligence is closely associated with the concept of Distributed Intelligence; and both should be considered twin sides of the same coin; namely constructive application of AI to the IoT problem space.

Situated Intelligence is when we have embedded Intelligence and Connectivity into the environment in useful ways. The result is a real-world object/system that gains the ability to sense, calculate, process and decide in relation to a specific class of problem-space; whereby aspects of the world (aka Things) suddenly ‘wakeup’ and begin engaging and interacting with people—and the environment—in a highly useful manner.

Situated Intelligence is what the IOT is all about—namely providing enhanced problem solving capabilities: including solution AUTONOMY (task replacement/management) and solution VISIBILITY (task awareness) with respect to Smart Things—or instrumented Devices, Systems and People—located at singular geographic LOCATIONS (specific IoT environments).

However, before we can attempt to discuss Situated Intelligence and what an IoT Science is, could or should be; we must first establish a clear concept of our primary subject matter. Henceforth we ask—what is the nature—and purpose—of the Smart Things that are spoken of specifically in relation to the IoT?

In order to find the answer, we must first consider the nature of technology—asking what it is—and where does it come from?


Nature of Technology

Noted technology experts and historians of science have given many and varied explanations of what technology actually is—and in terms of its most essential components/attributes – whereby they have attempted to define its most fundamental purpose or nature. Many have asked related questions—such as—why and how does a technology come to be?

In this respect, some people ascribe to an evolutionary theory—whereby new technologies are said to gradually evolve (semi-automatically?) from past developments, bolstered by improving human knowledge and manufacturing capabilities etc. Other commentators have stressed the key role that an inventor has to play in such a process.

Patently new technologies evolve or develop and they do become more sophisticated over time—and thus are rendered capable of—that is they enable—more complex capabilities as civilisation moves ahead. Typically newer technologies are comprised of combinations and assemblies of other (often older) technologies.

Technologies tend to appear as ‘bundles’ of other technologies—whereby one technology —over time—tends to becomes an atomic element—or component—inside another larger and more complex technology (in terms of both constructive elements and solution capabilities). In this way newer technologies ‘consume’ and are comprised of older ones.

This ‘bundling’ and evolution of technologies is not always the case (for example the hammer has remained relatively constant throughout time—a wooden handle attached to a heavy stone or metal head).  But often it is true that before a new technology emerges—that one or more older technologies which must have previously come onto the scene. For example, in order to develop the computer, human kind had to develop countless thousands of other technologies such as binary mathematics, manufacturing techniques for exotic materials such as silicon, transistors, microchips and microprocessors, plus memory chips etc.

Patently when we speak of the Internet of Things (IoT)—we are dealing with a vast number and range of different technology types, components and complex assemblies of sub-technologies; which are brought together for a huge variety of different reasons. But if this is so—what if anything—do all of these IoT technologies have in common?

It is salient to examine what are reasons behind the development of the IoT as a topic in the first place—and so to ask—what is the purpose of the IoT?


IoT Purpose(s)

We shall not examine—or argue—about how technologies come into being; but rather our focus shall be on the underlying goals/purposes of technology and how they can be achieved. What we are considering here, are the psychic, mental and sociological consequences of technology as it pertains to our individual and collective problem solving capabilities.

In other words, how does a technology amplify and/or accelerate existing human-managed processes (mental and physical); plus enable and/or extend new examples of the same.

Henceforth, the purposes of technology are to :

  1. Amplify / accelerate a state-of-affairs.
  2. Enable a state-of-affairs.
  3. Extend a state-of-affairs.

Whereby we define the meaning of a state-of-affairs to be a mental and/or physical state/process and/or product outcomes. One could say that the improved state-of-affairs relates to enhanced problem solving ability as detailed in the list of service aspects below.

Problem solving ability enhancement (service provision):

  1. Efficient service = faster workflows—timescale of task completion (pace).
  2. Magnified service = effective workflows—greater impact of task (scale).
  3. Reliable service = measurable workflows (reduced task uncertainty).
  4. Safe service = secure workflows—no negative consequences of task (safety).
  5. Automated service = automatic workflows (reduced human supervision).
  6. Economical service = cost-effective workflows (energy, environmental impact).

Henceforth, and put simply, the key attribute that a good or useful technology has—is that it produces a change of scale or pace, and/or enables a new pattern that is introduced into human affairs such as improved processes and/or enhanced knowledge (ref. mental and/or physical procedures etc). Technology accelerates and enlarges the scale of previous human functions, creating totally new kinds of cities and new kinds of work and leisure.

Put simply, here on this site, we shall define an IoT  technology as something which enhances the efficiency and/or effectiveness of human problem solving capabilities.

All in all,  and to be practically useful, the application of any useful technology must surpass the alternative(s) in one or more of the aforementioned respects; wherein the technology itself affords the solution of one or problems according to a set of task-specific efficiency factors.

Ergo, problems that may be solved by a properly applied IoT technology are:

  1. Pace: Results achieved in a shorter timespan (pace);
  2. Scale: Results achieved on new scales (physical/geographic);
  3. Pattern: New patterns are imbued into human affairs (mental/spatial/procedural);
  4. Energy: One or more of above are achieved with less effort/energy.

Goals, Intentions and Actions

A human-made Thing (Object/Tool/System/Machine)—constitutes a kind of extension of the human being, a manifestation of his physical, mental and psychical constitution.

In relation to the Internet Of Things or IoT we are normally speaking of a Thing that has been designed to provide a practical problem solving capability to one or more human beings—that is to get something done or else to perform useful Actions in the real-world. In other words we are talking of a technology that is created with a specific goal in mind; that is; the technology or Thing has a design purpose.

Often we can speak of the human as a tool-user who works in combination with the Thing or technology to achieve a specific goal or real-world state-of-affairs. This is achieved by means of an appropriate problem-solution model. Whereby we must start with some notion of what is wanted—and thus adequately define the Goal to be achieved. Next in order to meet a particular Goal any ‘helper’ Thing/Actor must interact with the world in such a manner as to foster its attainment (either autonomously or with human supervision).

We can speak of 4 different aspects of goal-achievement to consider:

  1. Goal: The Goal (what we want to happen in the real world);
  2. Action: What is done in the World (the Action);
  3. World: How it affects the outcome—possibly external to model;
  4. Evaluate: Check of the World.

Luckily famous design guru Dr Donald Norman has made an exhaustive study of the nature of Goals and how Things may be designed for efficient purposes (see his book ‘The Design Of Everyday Things’ ).

Donald has discovered that the Action has two aspects:

  • Execution: What we do to the World.
  • Evaluation: Comparing what happened with what we wanted to happen.

But real tasks are not so simple, because the original goal may be imprecisely specified, perhaps just ‘clean up the front-room’ is the top level goal—whereupon the Actor (a human or robot Hoover) must navigate a complex and changing world that is full of many obstacles and (potentially) blocking processes etc.

Goals do not expressly state what to do—or how to do it. In order to lead to appropriate action a goal must be transformed into specific statements of what is to be done—and Donald calls these intentions. An intention is a specific action taken to get to the goal.


Seven Stages of Action

Yet even intentions are not specific enough to control actions in such a manner as to successfully meet a goal. In fact there are seven stages of action that must be performed to achieve a Goal as follows:

  • Forming the Goal
  • Forming the Intention
  • Specifying the Action
  • Executing the Action
  • Perceiving the State of the World
  • Interpreting the State of the World
  • Evaluating the Outcome

These 7 stages form a problem-solution model of how to act to achieve a solution to a particular problem that needs to be solved;  however the stages are almost certainly not discrete entities; and most behaviour does not require going through all stages in sequence (on every occasion).

Most activities will not be satisfied by a single action (for example). There must be numerous sequences, and the top-level activities could last hours or even days (for example). There is a continual feedback loop—in which the results of one activity are used to direct further ones; in which goals lead to subgoals, intentions and sub-intentions etc.  Likewise there are activities in which goals are forgotten, discarded or super-seeded.

In conclusion, Norman’s analysis is highly illuminating—and because it draws our attention to key aspects of problem solving as a process which is fraught with difficulty that arises from basically 2 sources. Once a Goal is identified—firstly we have the problem of specifying and executing the supposed intentions and actions that may lead to problem solution; but secondly we have the problem of evaluating the results of actions. In any case, everything is dependant upon adequate specification of the problem context.


Problem Context

As stated the IoT is concerned with solving real-world problems—or (typically) with physical action aimed at the achievement of practical goals. Now with respect to the kinds of problems that arise in the physical world—a problem context has three aspects:

  • VISIBILITYtask awareness (problem/goal/solution specification)—create a visible mapping between the problem and solution space(s); form goal/intention/action facets accurately and precisely;
  • LOCATIONtask enhancement (resource allocation)—deal with constraints of location: perform key functions at a single or distributed location(s) and at the correct magnitude/scale and/or at specific multiple(s) of the same;
  • AUTONOMYtask replacement (resource management)—deal with system control/management/power overheads: autonomously or with an element of external support.

Now all three aspects of the problem context will involve specific challenges in terms of solution design and execution, and once again Donald Norman has indicated key aspects of a useful model.


Gulf of Execution and Gulf of Evaluation

In actual fact the world is a vast living process; and to be successful any Actor must constantly asses and re-sasses changing requirements—of essentially a constantly new situation.

One way of accurately modelling the world including all of this complexity/uncertainty is to admit, right away, that the difficulty arises in deriving the relationships between desired intentions and interpretations of the physical actions and states. There are several gulfs that separate modelling (or mental) states from physical ones —and these are to be avoided in the design of Things that are to be used by humans and/or that operate automatically on behalf of humans.

Firstly we have the Gulf of Execution—which asks—does the system provide actions that correspond to the Goal? The difference between intentions and the allowable actions is the Gulf of Execution. One measure of this gulf is how well the system allows the Actor to do the intended actions directly, without extra effort.

Secondly we have the Gulf of Evaluation—which asks—does the system provide a physical representation that can be directly perceived and that is directly interpretable in terms of the intentions and expectations of the Actor. The Gulf of Evaluation reflects the amount of effort that the Actor must exert to interpret the physical state of the system and to determine how well intentions have been met.

Now the Actor in question may be a human being operating a Thing—or the Actor (or Human) may be an automated Thing acting alone—or a combination of the two. Henceforth, these Gulfs of—Execution and Evaluation—will vary according to the problem context—and also with respect to the solution design (the problem solving Actor’s status and capability etc).


Solution Design: Classes of Action

Patently once we have specified the problem context sufficiently, we must turn our attention to solution design.

Now when considering the 7 stages of goal achievement; one would expect that the human designer of a Thing would normally play a large part in specifying the Goal, Intention(s) and appropriate Action(s) to attain the same. However when speaking of an IoT system we are typically concerned with imbuing the Thing  (essentially a man-made Actor) with a certain degree of intelligence and hence decision making plus self-automation with respect to the problem-solution process. Essentially this is a procedure that necessarily involves embedding a computer chip into the object to foster the same.

Perhaps the key attribute of a computer is that it can be programmed to perform a wide range of complex and highly specific tasks—and so to solve problems with a high degree of precision and accuracy.

It is salient  therefore to examine human-made Things in terms of the key concept of programability; whereby the OED defines program as:  “A definite plan or scheme of any intended proceedings; an outline or plan of something to be done… and as.. a sequence of objects, scenes, or events intended… a finally—a sequence of operations that a machine can be set to perform automatically.”

Henceforth, IoT objects/systems can be split into four kinds – according to the particular type of programmability enabled (or active capability) , as follows:

  • INERT THING—Controlled Action: provides particular functional abilities with respect to a specific context of use; an Inert Thing has no self-generated power of action, motion, or resistance; must be operated under external power and/or control; whereby such a device has no decision making ability whatsoever;
  • AUTOMATED THING—Automated Action = sensing/action/feedback loop (fixed problem-solution model): always responds to the environment in terms of a fixed pattern of predetermined  behaviours according to specific pattern of events; reactive in nature and has purely reflexive decision making (may be internally powered by some means);
  • PROGRAMMABLE THING—Programmable Action = sensing/action/feedback loop (programmable problem-solution model): responds to the environment in an appropriate way given a particular scenario; has pre-programmed decision making;
  • INTELLIGENT THING—Intelligent Action = sensing/action/feedback loop (adaptive problem-solution model): responds to the environment in an intelligent manner; has an ability to consider a range of options and/or allows for a scope of future ‘possibilities’ to be taken into account; can to adapt its own programming to new situations via learning behaviours; whereby said device has has pro-active decision making.

Now an IoT object will be not merely intelligent in terms of Actions provided; but will be capable of remote operation, environmental sensing, environmental integration and henceforth remote communication over a communication network etc: thus often connectivity will be provided to support associated functionality (e.g. Internet, WAN, LAN, WiFi,  Bluetooth, Radio connectivity etc);

IoT Thing/Object – connectivity types are as follows:

  • CONTROL CONNECTIVITY (Remote access/control/programming capability)
  • MONITORING CONNECTIVITY (Environmental sensing data)
  • INTEGRATION CONNECTIVITY (Collaborative operation data)
  • HOUSEKEEPING CONNECTIVITY (System status reporting data)

Plus the IoT object may be connected to environmental sensors (cameras, movement sensors etc) which may be provided and integrated into the overall IoT object/system (locally and/or remotely).

Now the IoT is concerned largely with Automated, Programmable and Intelligent Actions—and henceforth with respect to the aforementioned types of Things; each of which effectively prolong human problem solving abilities in the described ways. However when considering a solution space—programmability of Action is not the only dimension of interest; because as previously stated we are concerned also with the specific problem context in which said Action takes place (and associated intelligent responses).


Situated Intelligence

As stated, the problem context has 3 aspects: Visibility, Location and Autonomy; whereby we can define a solution in terms of Situational Intelligence—which encapsulates a Thing’s combined ability to provide a solution within a specific context; or in other words to: sense, cooperate, process/model, calculate, decide,  plus act upon environmental data etc;

SITUATIONAL INTELLIGENCE is defined as the ability of an IoT Thing/System to provide the following functional qualities (problem solving facets):

  • MONITOR (ENVIRONMENTAL SENSING):  degree to which an IoT object/system usefully SENSES data in relation to the environment;
  • INTEGRATE (PEOPLE, THINGS , SERVICES): degree to which an IoT object/system COOPERATES usefully with the environment;
  • ACT (MODEL, REPORT, CONTROL): degree to which an IoT object/system INTERACTS usefully with the environment.

Our new theory of IoT is grounded on a set of core principles, including for example the IoT Situated Intelligence (IOTSI) diagram (see below).

IOT_2C

SIOTF 1.0 – InfoGraphic A

IOT Situated Intelligence

Source: ‘The Science Of IoT’ (2020) – by Alan Radley

The IOTSI diagram teaches that in order for an IoT THING (or IoT Element)—existing within an IOT Environment—to provide a problem-solving capability in relation to a particular GOAL —then that same IoT THING must possess SITUATED INTELLIGENCE comprising a means of MONITORING, INTEGRATING and ACTING appropriately within said IoT Environment.

The IoT Environment consists of PEOPLE, THINGS and SERVICES.

The GOAL of an IoT SYSTEM (often comprising multiple connected IoT Elements) is to sense/know: Who/What needs Which capability, When, Where and How within this IoT Environment—and hence to provide the same. Whereby the IoT SYSTEM affords problem solving capabilities as follows: AUTONOMY (task replacement) plus VISIBILITY (task awareness) at a particular LOCATION (task enhancement). The IOTSA diagram indicates how a set of external GOALS (ref. MONITOR, INTEGRATE, ACT) are met using resources available to the IoT SYSTEM.

Patently an IoT System typically consists of a network of communicating IoT THING(S), the same ‘intelligent’ items connected together on a computer network; and each being capable of independently programmable yet coordinated behaviours, whereby SITUATED INTELLIGENCE is applied to meet the overall GOAL(s) of said IoT SYSTEM. Henceforth, SITUATED INTELLIGENCE refers to an IoT THING and/or group of IoT THINGS (an IoT SYSTEM) providing—and/or contributing to—an appropriate problem solving capability – aligned with one or more GOAL(S) (plus associated Tasks) – at a required geographical location and/or across a distributed set of geographical location(s).

The IOTSA diagram indicates how a set of system-wide Goals (ref. MONITOR, INTEGRATE, ACT) are met using resources available to the IoT System. Patently an IoT System typically consists of a network of communicating IoT Things, the same ‘intelligent’ items being connected together on a computer network; and each one is capable of independently programmable yet coordinated behaviours, whereby Situated Intelligence is applied to meet the goal(s) of said IoT System.

Situated Intelligence is what the IOT is all about—namely providing enhanced problem solving capabilities: including solution AUTONOMY (task replacement) and solution VISIBILITY (task awareness) with respect to Things—instrumented Devices, Systems and People—located at disparate geographic LOCATIONS.

Ergo, the Internet of Things (IOT) is a nexus of devices and services that allow for data exchange no matter where the end-user (human and/or machine) may be located. The author calls this ‘Situated Intelligence”—whereby a smart device knows WHO/WHAT needs WHICH capability WHERE, WHEN and HOW. Note that the Who/What (the requesting actor), Which (requested data/capability), When (normally ASAP) are normally known—and all that remains is WHERE AND HOW.


Problem Solving Ability

Notice that the concept of Situational Intelligence (aka Smartness) strongly relates to the concept of usefulness—here defined as the ability or quality to bring about good, advantage or benefit in a particular situational context;  whereby a Thing possesses appropriate functional qualities that relate to a specific purpose; whereby said object is serviceable to achieve a particular Goal.

Importantly Situational Intelligence (SI) relates to the purpose or Goal for which an IoT system has been designed or is being applied/used for a specific practical function. Whereby the Thing normally also presents itself as a summing-up, of said functionality; and has external (operative) aspects plus internal operational aspects.

An object imbued with Situated Intelligence possesses an ACTIVE aspect—which:

  • POTENTIATES (enables new capabilities—automation): Extends Operative Faculty of Human(s)
  • PROLONGS (current capabilities—available at new locations): Enhances Operative Faculty of Human(s)
  • MAGNIFIES (human powers—at new scale(s) etc): Amplifies Operative Faculty of Human(s)

Situated Intelligence is when we have embedded Intelligence and Connectivity into the environment in useful ways. The result is a real-world object/system that gains the ability to calculate, process and decide in relation to a specific class of problem-space; whereby aspects of the world suddenly ‘wakeup’ and begin looking-at and interacting with the environment in a highly useful manner.

The upshot is that each IOT Thing is imbued with situated intelligence for a distinct context-of-use. The result being—a sophisticated problem solving capability— that inherently includes both geographical and/or contextual knowledge plus an ability to perform targeted context-relevant actions etc.

A Smart Thing is an object can  sense (Monitor) and actuate/react (Act)  plus engage with objects/states/processes (Integrate) with respect to its environment (consisting of other People, Things/Systems [smart or stand-alone], and Services etc). In order to interact with its environment—in a useful/helpful manner—a Smart Thing must adequately interpret and securely process the information it collects, whilst protecting itself (and humans) from from threats and intrusions, plus communicating the results to other smart objects, people and systems while at the same time managing its own power consumption.

And it may be that the instrumented—or computerised/networked—elements of an IoT System and/or IoT Thing recede into the background of life. Whereupon we will have IoT Things scattered about – and these intelligent Things just know what they should be doing at any moment in time—and they just solve problems that arise of their own volition.

In other words, and as we stated earlier, IoT Things and Systems: Amplify, Enable and Extend human problem solving abilities. But the question remains as to how we should design an IoT System to adequately recognise and analyse problems, and to calculate, network and create/change a situation towards positive ends.

In other words, how do we design quality of service into the IoT?


Quality of Service

An IoT Thing (or IoT System) is designed to provide a real-world problem solving capability – a goal which often relates directly to an embodied place—or situational context—in the physical world. Location (X,Y,Z position), scale (physical dimensions—size) and responsiveness (temporal scale) are three key attributes in this respect.

Firstly in terms of location, individual IOT devices would typically be geographically located very close to the point at which the problem manifests—in situ—so-to-speak (or at least the IoT device would be able to remotely sense desired remote and/or extensive/pinpointed regions of the environment). In other words both problem solving recognition (monitoring facets) and problem solving means (solution facets) are tied closely together (i.e are either adjacent geographically, contextually coupled, and/or linked together by means of a network system).

Secondly, the issue of scale is crucial to the modern IoT vision—whereby the situated intelligence—or problem solving capability—is applied at the correct spatial size, and/or temporal timespan, plus at the correct magnitude (e.g. number of locations).

In terms of magnitude, we may have large numbers of connected objects scattered about the environment on tiny or large scales (for example). Some examples in this respect are street lighting control systems, wind farm turbines with individual adjustments, cellular pressure sensors, multiple heating elements, groups of sensors etc. Thirdly, the issue of temporal scale (typically including system responsiveness) is a vital aspect; whereby the problem solution should normally be provided within an adequate—or minimum— timescale.

Henceforth, as stated earlier, the key attribute that a good or useful IoT system has—is that it produces a change of scale or pace, and/or enables a new pattern that is introduced into human affairs such as improved processes/knowledge (mental and/or physical procedures etc).

Ergo, the IoT service would allow task-specific problem solution benefits as follows:

  1. Pace: Results achieved in a shorter timespan (pace);
  2. Scale: Results achieved on new scales (physical/geographic);
  3. Pattern: New patterns are imbued into human affairs (mental/spatial/procedural);
  4. Energy: One or more of above are achieved with less effort/energy.

In summary, we can measure the quality of service provision—for an IoT Service/Thing—by examination of how well the system provides and/or enables specific task-completion benefits—which carry over productively to the wider system Goal(s).

Henceforth, the productivity of an IoT System can be established and measured using generalised Quality of Service factors for an IoT System follows:

  1. Efficient service = faster workflows—reduced timescale of task (pace).
  2. Magnified service = effective workflows—greater impact of task (scale).
  3. Reliable service = accurate and reliable service (predictable results).
  4. Safe Service = secure workflows—free from negative consequences (safety).
  5. Automated service = efficient workflows (reduced supervision of task).
  6. Economical service = cost-effective workflows (energy, environment impact).

Conclusion

A new world of Programmable Things is on the horizon; whereby objects/systems are designed to possess embedded intelligence and connectivity—and henceforth an ability to calculate, process and analyse sensed information—whereby they can decide what form to adopt, plus functions and actions to exhibit or act; and in order to achieve a specified Goal.

In this section we have examined salient features of a generalised IoT system; whereby we have demonstrated how a particular IoT problem context is matched to its specific IoT solution space using the new concept of Situated Intelligence.

Along the way we have identified 3 key aspects that all problem context(s) have in common; namely: Visibility, Location and Autonomy of problem context. Plus we have described 4 classes of IoT Things (Inert, Automated, Programable and Intelligent) that may be employed to address said problem contextual aspects. Finally we have delineated 3 functional problem solving capabilities; namely the ability to: Monitor, Integrate and Act appropriately within the IoT environment.

In sum, it is hoped that this new analytical framework can enable a close mapping between problem and solution spaces—leading to IoT System designs that perform optimally and in predictable ways. All in all,  we have demonstrated how a set of generalised Quality Of Service factors can be used to ensure that the IoT System is designed and built to meet operational requirements in full.