duminică, 22 octombrie 2017


Acknowledgements: The author wants to thank XpertRule Software LTD and mr. Tim Sell for being able to try Decision Author- the main software in which this prototype shall be built.


The design of the expert system starts by defining heuristic knowledge regarding the aspects that would be assessed. The primary parts are named Questions. Xpert Rule offers a variety of question types, from the simpler yes-no to more complex with multiple answers, etc.
The knowledge structured for the primary knowledge (as yes/no questions) is presented in figure 1.

Figure 4- Knowledge structured in yes/no questions

Figure 5-Primary and mid-level knowledge structured in questions

Figure 6- All the knowledge is structured here

The schema of structuring knowledge is presented in the figure  below.

Figure 7 Structuring Knowledge


Xpert Rules is configuring the expertise required to solve the problem around a decision tree. We have constructed our decision thee as in the figure- in order to:
-catch as much as possible of primary knowledge- that is main heuristic knowledge;
-use mid-level knowledge- that is the knowledge managed generally by supervisors and floor level managers- as needed;
-use also high level knowledge;
As our main goal is to make a risk/safety assessment- all the acquired/elicited knowledge was tailored in this respect- as seen in figure 8

Figure 8 Knowledge coupling for the building of the decision tree

Actually there are three distinct trees in the structure:
a. The Start Tree is the default Xpert Rule structure. In order to execute the expert system all the other trees must be copied/assembled into the Start Tree.
b. The EMPLOYEE_SAFETY_TREE is the main tree of this expert system- with the knowledge organised as presented before into three zones of primary, mid-level and high.
c. Tree_cont is the continuation of the main tree with the two procedures that are defined to count the Risk/Safety level and to evaluate it. This tree was developed in the idea to modularise the structure- otherwise the procedures could be assembled into the EMPLOEE tree.
Figure 9 shows the hierarchy of these trees as appears in the design window.

Figure 9 Knowledge trees

The knowledge maps of the EMPLOEE and Start trees are shown in the next figures.
Figure 10  Knowledge map of EMPLOEE_SAFETY_TREE

Figure 11 Knowledge map for the Start Tree

Structures of the primary knowledge part, of the mid- level knowledge part and of the high level knowledge part are presented in the next figures.

Figure 12 Primary Knowledge part of EMPLOYEE_SAFETY_TREE

Figure 13 Mid-level Knowledge part of the same tree

Figure 14 High level Knowledge

Figure 15 Start Tree

Here are collected some data regarding the audit- as auditor name and the place where the audit was performed.

Some aspects from a trial run of the expert system are shown in the figures below.
Figure 16 Yes/No Question

Figure 17 1-3 scale question

Figure 18 1-5 scale question

Figure 19 Partial solution


 Could be such an approach helpful for the safety and health domain ? The development and upgrade of specialised expert systems could transform the process of safety assurance, especially in process industries into a  facile and efficient matter- as the existing heuristic knowledge could be elicited and added to the already existent knowledge. The development of an expert system is no more a very specific IT problem- such systems could be developed also by non- IT specialists.
We are interested to pilot our experience to interested parties.  


Bech-Larsen, T. and Nielsen, N.A.(1999) A comparison of five elicitation techniques for elicitation of attributes of low involvement products. Journal of Economic Psychology 20, 3 (1999), 315-341

Grunert, K.G. and Bech-Larsen, T.(2005), Explaining choice option attractiveness by beliefs elicited by the laddering method. Journal of Economic Psychology 26, 2 (2005), 223-241.

Jans, G. and Calvi, L.(2006) Using Laddering and Association Techniques to Develop a User-Friendly Mobile (City) Application. In On the Move to Meaningful Internet Systems 2006: OTM 2006 Workshops. 2006, 1956-1965
Reynolds, T. and Gutman, J.(1988) Laddering theory, method, analysis, and interpretation. Journal of Advertising Research 28 (1), pp. 11-31, 1988.

Vanden Abeele, V. and Zaman, B.(2008), The Extended Likeability Framework: A Theoretical Framework for and a Practical Case of Designing Likeable Media Applications for Pre-schoolers. Advances in Human Computer Interaction 2008


Acknowledgements: The author wants to thank XpertRule Software LTD and mr. Tim Sell for being able to try Decision Author- the main software in which this prototype shall be built.


Safety domain of research is by excellence a domain based on expertise. Textbooks and theoretical knowledge are good but the safety expert which inspects three times a day a certain part of an enterprise is the ultimate safety dealer here.
A lot of expertise is transformed into lessons learned- that are used for training and improvement of existing safety attitudes. On the other part, this expertise could be also valued in order to build optimal and effective safety assessment systems.
An expert system is software that emulates the decision-making ability of a human expert. In our case- the expert part should interrogate the specific employees regarding safety aspects of an enterprise.
The next figure illustrates how a safety expert, with the necessary knowledge into the problem could improve a safety assessment process. The safety manager of a specific enterprise which performs a safety assessment will navigate through the safety assessment items, will assess something (the current safety situation inside his enterprise- preferably) and will develop a safety improvement plan- this being the outcome of the assessment. of course that this assessment could be external and then we will have third party auditors- but essentially the process is the same. We need from our safety expert:
-to design optimally and efficiently the safety assessment items- so that the answers of the safety manager should capture the real, most significant and important from the safety point of view aspects;
-to develop the objective and optimal safety assessment procedure so that a very subjective process should perform as objective as possible;
-to build a reliable and cost sensitive improvement plan that will do exactly this- to improve safety and health on the basis of the performed safety assessment.
Generally, the existing safety assessment methods are using textbook based evaluation items- in order to assure a very large usage possibility. This is good- on one part- because they are giving us a certain referential that is the same in USA, France or Romania. This is bad, on another part- because it is not using the existing safety expertise in defining and developing the safety assessment items.
In our paper we are proposing an original approach regarding the development of optimal safety assessment solutions using expert system developments.

Figure 1 Safety Expert helps Safety Manager to perform an objective assessment

Safety Knowledge Elicitation

Ok, we have tons of safety knowledge. Do we need more? What is in into knowledge elicitation? Expertise is a collection of knowledge, experience, skills, techniques, facts, rules and so on, and using them to lead to goals. This is also valuable for safety expertise. We must say that expertise is unique reported to the context, to the domain and also to the place where expertise is established. The knowledge is elicited chiefly from experts in the field and data/ information available from published literature. There are various known knowledge elicitation techniques available. The choice of technique to be used in the knowledge elicitation process depends on the nature of the situation within which the knowledge is elicited, the domain knowledge and availability of experts. Elicitation could be defined as the process of acquiring knowledge from a domain expert- with some specific problems. The knowledge engineer (Bech-Larsen, 1999) will find it difficult to structure the interviews. On one hand interrupting the expert may break his train of thought and lose useful information; on the other hand, along digression into anecdotes or irrelevant information will make the transcript difficult to understand. However, it is hard to judge what is irrelevant, or where the expert train of thought is leading. The problems are compounded by the shades of colour in language. The expert will use technical terms, and the knowledge engineer must make sure that he and the expert have the same understanding of this terminology. The words such as big, small, frequently, sometimes are fuzzy, it is difficult to define. Also, the definition may depend on the context. It may be that the expert himself is not even consistent in his use of words (Vanden, 2008).
Laddering (Reynolds, 1988) is a structured questioning technique derived from the repertory grid technique, a, specific knowledge elicitation method that could apply well into safety. In Laddering, a hierarchical structure of the domain is formed by asking questions designed to move up, down, and across the hierarchy... Laddering is enabling a hierarchy of concepts to be established. The common generic means-end chain, therefore, consists of Safety Attributes (SA), Safety Consequences (C) and Safety Added Value (SAV).
Safety Attributes -> Safety Consequences -> Safety Added Value
Laddering (Grunnert, 2005) allows crossing over from qualitative data gathering to quantitative data treatment. The laddering data analysis process (Jans 2006) typically involves the following two phases. First, the laddering interviews are transcribed and core elements -attributes, consequences, and values- are coded. This phase relies on the qualitative research tradition, requiring skills such as axial coding to define elements that emerge from the interviews. Once that the core elements are defined and labelled, the individual ladders can be decomposed based on these codes.
Laddering is illustrated in the next figure.

Figure 2 Laddering as safety knowledge elicitation technique

20 Questions is a method used to determine how the expert gathers information by having the expert as the knowledge engineer questions- having as output a. amount and type of information used to solve problems; how problem space is organized, or how expert has represented task-relevant knowledge.
In our research we have built a questionnaire for identifying, capturing and using heuristic knowledge, considering that we could add significant amount of knowledge to the existing one in the already done assessment. An example is given below.
New Knowledge elicited
Elicitation trigger
Former Knowledge
You must check especially if a maintenance job is repetitive done on the pipes that are alimenting processor X. Maintenance teams are usually solving punctual problems- without checking the safety of the whole structure. Be especially careful if they used to heat the pipes in order to make the maintenance.
From your experience- could you tell us something that could jeopardise the safety regarding alimentation pipes for processor X?
Alimentation pipes shall be checked.

Decision Author is well suited in order to introduce”Knowledge on the fly” into the expert structure, knowledge derived from elicitation.
The elicitation process is presented in detail in the next figure.

Figure 3 Elicitation process

To be continued...