In category Articles

A well-designed industrial diary could provide an essential tool for intelligent technology

Antero Karvonen, Olli Kuusisto & Pertti Saariluoma

Knowledge is a logical requirement for autonomous and intelligent technology to perform their functions. In terms of current, especially brownfield, industrial processes this presents a problem, given that much of that knowledge is in the human domain; it is largely tacit (1.) and shared informally. As a continuation of earlier work on a ’Next-generation Diary (2.)’ , we argue that an intelligent industrial diary could act as a bridge and interface between machine and human intelligence.

This text is based on a forthcoming article to be presented at the 5th International Conference on Intelligent Human Systems Integration: Integrating People and Intelligent Systems (IHSI 2022) to be held in Venice, Italy, February 22-24, 2022.

Design of intelligent technology and cognitive mimetics

On the frontier of future industry, the question of intelligent technology looms large. One basic question has to do with approach: how should machine intelligence be designed. This presents are myriad of technical and pragmatic questions, one of which is the basis and method by which a computational system is aligned or harmonized with real processes. The way in which human operators cope and maintain grip over the processes is through skills and domain- and general knowledge. In our view, the cognitive mimetic approach, which seeks to open these human information processes from a cognitive perspective and transfer them to technology, is a promising method for designing machine intelligence.

Human Digital Twins and Industry 5.0

Digitalization of industrial processes, characterized by e.g. connectivity, automation, machine learning and real-time data, is often referred to as industry 4.0. Digital twins are a key example of the potential digitalization could offer. Earlier1, we have written of the idea of human digital twins (HDTs). Where digital twins are replicas of technical processes, HDTs are replicas of the (associated) human information processes. Hence, they are kinds of models of humans interacting with industrial processes, based on the cognitive mimetic approach. HDTs are thematically at the leading edge of so-called industry 5.0, where increasingly sophisticated technology is combined with an increasing focus on the human-technology system.

Intelligent industrial diaries

By combining the ideas of cognitive mimetics and human digital twins with the knowledge requirements for intelligent technology and the idea of an intelligent industrial diary we arrive at a promising confluence of ideas. By designing the diary mimetically, the promise is that we can solve two problems: match and fit of the diary with real human work, action and thought on one side, and match and fit the contents of the diary with computational control systems on the other.

One thing is to open information processing in human mind while they carry out some intelligence requiring task. Human knowledge is mostly tacit or subconscious, but it is possible to open this knowledge by the methods of cognitive science and psychology. It is possible to get an idea about how people process information when they carry out their tasks. Knowledge of human information processing may be used in constructing intelligent information technologies. The analysis of chess players’ thinking made it possible to construct chess machines which play better chess than any human being. Similarly, tasks carried today by human operators can be realized by AI in the future, if only it is known how people process information in those tasks. In terms of the industrial diary, these methodological ideas could become embodied in the operating principles and interaction patterns of the artefact itself. To reach this vision a number of technical and conceptual issues must be resolved. Next we will make a few remarks on these issues, focusing on ontologies, leaving many important issues for future articles.

Ontologies, patterns, and relations

One of the core themes in this mimetic endeavor is describing human information processes by means of an ontology. In our case, many of the features and user requirements surfaced during research on the concept of the future diary implicate ontologies, such as intelligent text recognition, structured but not compulsory entry-making, and intelligent search, to name a few. On the other hand, ontologies as structured knowledge logically provide also a basis for machine intelligence. In this way, the challenges of well-designed interaction and intelligent machine control can be solved at a common core of ontological engineering and iteration.

Ontologies are essentially representational structures, that are used to ’carve the world at its’ joints’. Ontologies can be constructed on multiple levels. So-called upper ontologies are built out of highly abstract concepts such as objects, relations, states, events, and processes. On ”lower” level, we have domain-ontologies that are used for structuring knowledge of a particular domain, such as a pulp mill. In a forthcoming conference article, we explicate these in more detail. For present purposes, some key elements of this upper ontology can be highlighted: states and actions which can be combined through various types of relations. The idea is to allow the representation of chains of events and actions occurring at the mill.

While work on upper ontologies is necessary for system design, we should not lose sight of domain-ontologies that are the real target of intelligent knowledge capture. On this level, the general and highly abstract concepts become more specified and domain-specific: types of events and actions, associated with concrete objects. The key element at this level we have called patterns. This is the dynamic and evolving level of the ontology, which should be both well-designed at the start by designers, but also subject to revision. In fact, we feel that the upper ontology should also be subject to revision, if empirical reality shows it to be necessary.

A pattern is essentially a recurring type of event-action pair that is populated with concrete contents or references to the actual mill. It stands in a relation to the domain-ontology in there can be no elements in a pattern not found in the ontology. On the other hand, no pattern is without reference to a set of entries of which it is an abstraction. Thus, patterns are essentially prototypical, constructed in a way that has explicit reference to concrete occurrent events and the domain-ontology. Space does not permit going into detail, but future research will involve the processes by which patterns and thereby domain-ontologies are constructed vis-a-vis interaction with the intelligent diary.


A key objective of an intelligent industrial diary is to accumulate employees’ tacit knowledge, combine it in a context-specific manner to other information, and process it so that information and experiences can be shared between shifts, departments, and factories/mills, and easily searched for and exploited also in relation to other systems and technologies in the future.

Conceptually, the ontologies, patterns, and relations sketched out here seem promising for establishing the basis for an intelligent diary that is mimetic with respect to empirical information processes in a pulp mill and that approximates the human side of the mill operations. While these remarks are promissory, nothing in them seems technically or theoretically unfeasible. For example, language technology and ontological techniques have advanced significantly, which we believe is an enabling factor for the capture of empirical domain ontologies through the concept of pattern. Through the concept of directed associations (of various types), these can be transformed into networks of chains of events. These, in turn, could lead to machine- tractable generalizations when combined with the pattern ontology and digital twins and thus form a basis for intelligent technology.