Preface

Context has always played an important, if little understood, role in human intelligence. This is especially true in human decision-making and communication. An individual's awareness of their own context as well as that of others with whom they interact allows many assumptions to be made about the discussion, the environment and/or the problem at hand. This allows many important aspects of human interaction to remain implicit when the communicants are in a common context, or alternatively, in different but mutually understood contexts. Otherwise, all assumptions would have to always be explicitly spelled out, which is a truly burdensome task for everyday communication.

Brézillon and Abu-Hakima (1995), in their report on the IJCAI-93 workshop "Using Knowledge in its Context", provide a summary of the discussions between the participants before (by electronic mail) and during the one-day workshop. Brézillon and Abu-Hakima clearly show that the notion of context was far from defined and is dependent in its interpretation on a cognitive science versus an engineering (or system building) point of view. The identification of these two viewpoints made it possible to go one step further than previous workshops. Once a distinction is made on the viewpoint, we can achieve a surprising consensus on the aspects of context, mainly the position, the elements, the representation and the use of context. However, despite the consensus on the aspects of context, agreement on the notion of context has not yet been achieved. Indeed, we still continue to see the two opposing views on context seen at the IJCAI-93 workshop. This book is an attempt to reconcile these two views.

The notion of context has been defined by a number of authors. Bazire and Brézillon (2005) presented a study on 166 definitions of context, but now we have 268 definitions. The results of their study show that context is specific to a given situation after answering the following questions: Who (i.e. the actor of the action)? What (i.e. the focus)? Where (i.e. the spatial location)? When (i.e. the temporal location)? Why (i.e. the intentions, the goals and even the emotions of the actor)? and How (i.e. the procedure needed to realize the action)? This study showed that the optimal order of presentation is action-object-agent-location, which gives a primary role to the action that confirms the necessity to consider the goal as a part of context. Webster's Dictionary defines context as "the whole situation, background or environment relevant to some happening or personality". This definition suggests that the context is always tacit and is rarely mentioned explicitly. Another definition of context, which is more operational in computing, is: "Context is what constrains a focus without intervening in it explicitly" (Brézillon and Pomerol 1999). Some researchers and practitioners have proposed singular and often narrow definitions, and focus on some aspects of context that can be identified from data, information or knowledge obtained through sensors that are specific to their domain or discipline (mainly around the context-aware computing community). In problem-solving, the context inherently contains much knowledge about the situation and environment of the problem. The context indeed constrains the focus, but conversely, the focus allows for the specification of the relevant contextual elements. For example, a dead battery in a car that has been parked overnight in freezing temperatures constrains the focus of a diagnostician, and it also specifies the contextual elements they must consider. Such a context has entirely different diagnostic implications than one where the car is in operation when the battery dies. Therefore, the effect of context on problem-solving and decision-making can likewise be very significant.

Context also introduces the expectation of a behavior that, by convention, goes along with a specific situation. Turner (1998) exemplifies this by alluding to the fact that when an individual enters a library, conversation is habitually reduced to a whisper. As the context changed (when the subject entered a library), new behaviors (whispers) were instantiated for use in the new context. Upon leaving a library, the individual's context changes again, and the prior behavior is no longer enforced.

Context affects virtually all aspects of behavior in animals, humans and computer systems. It affects how we understand the world, communicate with others, and plan and carry out our actions. It affects how computer systems should behave so that they act appropriately for their situation and their users.

We can find old references in some disciplines. For example, in mathematics, Frege (1892) identifies three different contexts: ordinary, direct and indirect contexts. Several theories also refer indirectly to context, such as activity theory (Leont'ev 1978), situation theory (Lave 1988) and distributed cognition (Flor and Hutchins 1991). It is generally acknowledged that John McCarthy brought the role of context in logic to prominence in 1993. The basic relation he introduced, ist(c,p), means the proposition p is true in the context c. In other words, "context captures all that is not explicit in p but is required to make p a meaningful statement". There are several important conclusions in McCarthy's paper:

• - a context is always relative to another context;
• - contexts have an infinite dimension;
• - contexts cannot be described completely;
• - when several contexts occur in a discussion, there is a common (shared) context above all of them into which all terms and predicates can be lifted.

Note that in this formal approach, context is supposed to have a discrete nature, while in a cognitive approach there is a unique context of interest - the context of communication and interaction - which evolves continuously with the interactions.

In other disciplines, researchers consider it pointless to directly model context (one argument is the infinite dimension of context) and an external approach is preferred. For example, knowledge engineers introduced screening clauses in rule-based systems based on external control knowledge (not called context). For example, Clancey (1983) presents the following rule of MYCIN:

IF

1. The infection is meningitis,
2. Only circumstantial evidence is available for this case,
3. The type of meningitis is bacterial,
4. The age of the patient is greater than 17 years old, and
5. The patient is an alcoholic,

THEN

Prescribe tetracycline.

The fourth clause has nothing to do with the identification of the meningitis, but rather avoids firing the rule for children because tetracycline discolors growing teeth. Here, there is a unique context of interest that is intricately linked to the reasoning.

Contextual effects have been studied in many different disciplines over the years but usually without too much interaction between researchers in the different fields. This began to change in a small way in the mid-1990s with the 1993 International Joint Conference on Artificial Intelligence on modeling context in knowledge representation and reasoning, which sought to bring together some of the various threads of context-related research. The interdisciplinary focus on context as a subject of research truly began in 1997, when the First International and Interdisciplinary Conference on Modeling and Using Context, CONTEXT 1997, was held in Rio de Janeiro.

The two co-editors of this book were at the origin of the CONTEXT conference series. Our idea was to have a real interdisciplinary conference covering any discipline concerned with modeling and use of context. Disciplines ranged from philosophy to application development. However, over the last 25 years, the rapid pace of new technology resulted in the increasing importance of applications, and theoretical aspects of context without immediate spin-offs were of little interest to developers. As a result, for the last four years, the focus of the CONTEXT series has moved toward the pragmatic aspects of modeling and using context in order to serve all researchers and practitioners who develop concrete solutions based on any conceptual framework. Thus, we expect the audience of the book to be primarily those interested in developing pragmatic, operational frameworks for problem-solving by intelligent systems, not necessarily those interested in the more formal aspects of modeling context.

Conferences, after Rio de Janeiro (1997), have been held in Trento, Italy (1999), Dundee, UK (2001), Stanford, California, USA (2003), Paris, France (2005), Roskilde, Denmark (2007), Karlsruhe, Germany (2011), Annecy, France (2013), Larnaca, Cyprus (2015), Paris, France (2017) and Trento, Italy (2019), with the next conference planned for 2023. As the Covid-19 pandemic continued throughout 2021, a special issue of Modeling and Using Context was created in place of the conference. The lessons learned are that: (1) researchers come from different disciplines and wish to share ideas with researchers in other disciplines for solving similar problems by using context; and (2) the two opposite views on context, identified in 1993, are always present. The engineering viewpoint (the "hard science" following McCarthy and Giunchiglia), for whom contexts are discrete objects, and the cognitive viewpoint (the "soft science" following Simon and Newell), for whom there is a unique...