[Paper Review] Overview and Challenges of Ambient Systems, Towards a Constructivist Approach to their Modelling

Gérald Rocher, Jean‐Yves Tigli+2 more|arXiv (Cornell University)|Jan 9, 2020

Context-Aware Activity Recognition Systems78 references4 citations

TL;DR

This paper proposes a constructivist systemic approach to modeling ambient systems—such as IoT, CPS, and ubiquitous computing—by shifting from predictive modeling to in vivo effectiveness evaluation. It argues that due to environmental complexity and unpredictability, traditional prediction fails, and instead, system effectiveness must be assessed empirically through purpose-driven models, enabling better adaptability, user experience, and dependability in real-world deployments.

ABSTRACT

From a closed and controlled environment, neglecting all the external disturbances, information processing systems are now exposed to the complexity and the aleas of the physical environment, open and uncontrolled. Indeed, as envisioned by Mark Weiser as early as 1991, the progresses made on wireless communications, energy storage and the miniaturization of computer components, made it possible the fusion of the physical and digital worlds. This fusion is embodied in a set of concepts such as Internet of Things, Pervasive Computing, Ubiquitous Computing, etc. From a synthesis of these different concepts, we show that beyond the simple collection of environmental data from sensors, the purpose of the information processing systems underlying these concepts is to carry out relevant actions that the processing of these data suggests in our environment. However, due to the complexity of these systems and the inability to predict the effects of their actions, the responsibility for these actions still often remains with users. Mark Weiser's vision of disappearing computing is still far from being a reality. This situation calls for an epistemological rupture that is proposed to be concretized through the systemic approach which finds its foundations in constructivism. It is no longer a question of predicting but of evaluating in vivo the effectiveness of these systems. The perspectives for such an approach are discussed.

Motivation & Objective

• Address the epistemological challenge of modeling complex ambient systems where predictive models fail due to environmental unpredictability.
• Overcome the limitations of traditional system modeling in open, dynamic environments where user responsibility for system actions remains unmanaged.
• Introduce a systemic modeling approach rooted in constructivism to evaluate system effectiveness in real-time, rather than relying on isomorphic prediction.
• Enable practical assessment of user experience, dependability, and adaptability in ambient systems through effectiveness-driven evaluation.
• Provide a foundation for auto-adaptive systems and robust design in cyber-physical environments by grounding evaluation in observable, real-world outcomes.

Proposed method

• Adopt a systemic modeling approach grounded in constructivism, rejecting isomorphic prediction in favor of empirical evaluation of system behavior.
• Model ambient systems based on their intended purposes, using a purpose-driven framework that captures functional goals in physical environments.
• Use probabilistic Input/Output Hidden Markov Models (IOHMM) to assess system effectiveness, with extensions into possibilistic frameworks to handle temporal constraints.
• Evaluate system performance not through isolated non-functional metrics (e.g., MTBF, MDT), but via high-level effectiveness indicators reflecting real-world impact.
• Integrate user experience and dependability into the evaluation by linking effectiveness to subjective quality of experience (QoE) and system trustworthiness.
• Apply the model in real environments to assess system behavior in vivo, avoiding reliance on controlled, synthetic test conditions.

Experimental results

Research questions

• RQ1 How can ambient systems be effectively modeled when their environmental interactions are inherently unpredictable and context-dependent?
• RQ2 What epistemological shift is required to move from predictive modeling to real-world evaluation of system behavior in complex, open environments?
• RQ3 To what extent can effectiveness assessment serve as a unifying metric for dependability, performance, and user experience in ambient systems?
• RQ4 How can systemic modeling based on constructivism support the development of auto-adaptive and context-aware systems?
• RQ5 Can effectiveness evaluation replace traditional non-functional metrics (e.g., MTBF, availability) as a holistic indicator of system trustworthiness?

Key findings

• Traditional predictive modeling is insufficient for ambient systems due to environmental complexity and the inability to anticipate all interactions.
• Effectiveness assessment, evaluated in real-world conditions, provides a more pragmatic and reliable alternative to isomorphic prediction.
• The systemic approach rooted in constructivism enables evaluation of system behavior based on purpose, rather than on theoretical models.
• Probabilistic IOHMM and possibilistic extensions allow effective modeling of system behavior under uncertainty and temporal constraints.
• Effectiveness assessment can serve as a high-level indicator for dependability, performance, and user experience, subsuming traditional non-functional metrics.
• First implementations show promise in supporting auto-adaptive systems and improving robustness in real-world ambient computing environments.