5. Process models and heuristics for mining HCI data

By the end of Section 3, the authors have summarized their HCI data mining approach and visualized essentials of inductive modeling in Figure 3. We take up the thread once again. The selection or the design of a terminology is essential. The terminology determines the space of hypothetical models that may be found. Throughout the process of data mining, model spaces may be subject to revision repeatedly (see preceding Section 4).

The world of models is overwhelmingly rich. Models may be characterized by properties, by purpose, by function, by model viability, or by model fitness [30]. As Thalheim puts it, "models are developed within a theory" ([30], p. 117).

Every concrete application domain provides such an underlying theory. It is a necessary precondition to data mining to specify all the aspects of the underlying theory that should be taken into account (see [30], p. 115, for mapping, truncation, distortion, and the like). Revisions may turn out to be necessary, when inductive modeling, i.e., learning proceeds. Therefore, the word "data understanding" in the CRISP model (see Figure 1) is considered inappropriate and, hence, substituted by "data analysis" in the approach shown in Figure 3. This figure is intended to visualize both the dynamics of the data and of the model spaces. Hypothetical data understanding is seen as the preliminary result of data mining.

Another question derives from the authors' generalization of identification by enumeration. The authors are convinced that it is possible to generalize their recent approach to dynamic identification by enumeration even further. This requires a careful easing of one or more of the requirements named operational appropriateness, conversational appropriateness, and

Mining HCI Data for Theory of Mind Induction http://dx.doi.org/10.5772/intechopen.74400 61

Finally, the authors want to attract the audience's attention to a larger and rather involved field

There are rarely any bug-free software systems. In the future, there will be rarely any bug-free assistant systems. However, even if a future assistant system were to be totally free of bugs, it would hardly be able to solve every imaginable problem. Digital assistant systems may fail. In response to this severe problem, it is necessary to work toward digital systems able to ponder

Limitations of learning systems are unavoidable [17]. In response, approaches to reflective inductive learning have been developed and investigated in much detail [75]. The results

The authors' step from the conventional approach to dynamic identification by enumeration reveals a feature of reflection. A learning digital assistant system that gives up a certain space of hypotheses—in formal terms, γ(f[n]) 6¼ γ(f[n + 1]) resp. γ(fX[n]) 6¼ γ(fX[n + 1])—with the intention to change or to extend the terminology in use is, in a certain sense, reflective. It "worries" about the limits of its current expressive power and aims at fixing the problem. Vice versa, a system able to change spaces of hypotheses, but not doing so (formally, it holds γ(f[n]) = γ(f[n + 1]) or γ(fX[n]) = γ(fX[n + 1]), resp.), shows a certain confidence in its abilities to solve

This leads immediately to a variety of possibilities to implement reflective system behavior. First, a system changing its space of hypotheses may inform the human user about its recent doubts as to the limitations of terminology. Second, a bit further, it may inform the human user about details of the new terminology. Third, such a system may also report confidence.

As a side effect, so to speak, the authors' work leads to concepts and algorithmic approaches to reflective AI. This bears strong evidence of the need for further in-depth investigations.

After the second author—inspired by some fascinating results in behavioral sciences has introduced the concept and coined the term of theory of mind modeling and induction in 2012, the two authors' student Bernd Schmidt has undertaken the endeavor to provide the first theory of mind modeling and induction application. The authors are grateful to him for his engaged and excellent work and for his continuous willingness to meet whatsoever

semantic appropriateness. The related open questions need some more research effort.

of research problems beyond the limits of this chapter: reflective artificial intelligence.

demonstrate the possibility to design and implement reflective artificial intelligence.

their own abilities and limitations. Systems that do so are called reflective.

the current problem.

Acknowledgements

requirements.

When speaking about logics and its algorithmic use, it is strictly advisable to stay within the limits of first-order predicate calculus [71]. The selection or the design of a logic means to decide about the signature of the language and about axiom sets of background knowledge.

Under the assumption of a given logic, business understanding and data analysis underpin an impression of what the current analysis process is about. To say it more practically, what might be typical statements arrived at by the end of the data mining process? In the authors' digital game case study, by way of illustration, typical statements explain a human player's action under conditions of a play state [4, 5]. In their business intelligence application [6–8], formulas relate business data and temporal information of largely varying granularity. As soon as the type of expected formulas becomes clear, the next design task is to specify an indexed family of logical formulas. This forms the first space of hypothetical models.

Within the authors' framework, a crucial step is the modification of a space of hypotheses. There are heuristics discussed in [8] that shall be briefly surveyed. An automation may require, to some extent, natural language processing.

The human user's activities are syntactically analyzed. In case there occur terms that have no corresponding sort, constant, function, or predicate names in the formulas of the current space of hypotheses, a limitation of the terminology is detected. The system is "unable to speak about what the user is doing." A case discussed in [8], p. 234, is "retracement of business volume." Retracement is interpreted as inequality with a (large) factor in it, and some sequence of such formulas of properly increasing strength is automatically generated.

Methodologies, guidelines, and process models aiming at (logical) model space construction are worth much more future research work and practical exploration.
