ATTENTION, INTERPRETING, DECISION-MAKING AND ACTING IN MANUAL ASSEMBLY. Gunnar Bäckstrand 1,2,3, Leo J De Vin 2, Dan Högberg 2, Keith Case 1,2 1. Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, Leicestershire, LE11 3TU, UK. 2. The School of Technology and Society, University of Skövde, Box 408, SE-541 28, Skövde, Sweden. 3. Volvo Powertrain AB, SE-541 87, Skövde, Sweden. ABSTRACT In a modern manufacturing environment, data and information are a vital part of the manufacturing process and in particular for supporting the value adding activities. Modern manufacturing information systems allow fast distribution of, and access to, data and information. However, the technical improvements of manufacturing information systems do not always create the benefits that were expected from them. This paper discusses this problem in the context of manual assembly tasks. Attention, interpretation and decision-making are important drivers for how well the assembly tasks are performed - the acting. In other words, the acting is governed by how and when the attention of the assembly operator is caught, how easily the information can be interpreted, and to what extent the information is useful for decision making. The aim with the work is to find and present why data and information provided on the shop floor often fails to prevent quality problems; not seldom this data and information actually causes the problems. This paper focuses on one of the core issues related to assembly data and information, namely active attention and how this is triggered. If active information seeking behaviour is not present on the assembly shop floor, then the probability for a quality problem increases. KEYWORDS: Manufacturing, Information Impact, Attention 1. INTRODUCTION Volvo Powertrain in Skövde, Sweden manufactures heavy diesel engines used in trucks, excavators, articulated haulers, boats etc. The engine assembly represents mixed model production. In the main flow, a high volume product is assembled, but there also low volume products assembled, intertwined with this product flow. The presence of these low volume products demands a dynamic information interface and a dynamic information flow concerning information transfer of information for individual engines. The character of this information can refer to for instance parts to be assembled on the specific engine, but it can also be information regarding how to assemble the engine. The aim with this paper is to try to identify explore the relationship between attention, interpretation, decision-making and acting and how this relates to the information flow. The work uses a hypothesis formulated in previous work [1] as a starting point. This hypothesis is: The degree to which Active Information Seeking behaviour is supported/triggered has a large influence on the number of internal rejects. As an extreme example, if a trigger is not present or detected, then an active information seeking behaviour will not be present. In other words, the trigger is predecessor to active
attention, and active attention is believed by the authors to be a state of mind that is crucial for successful use of data and information use. The focus in the paper is on attention because this is believed to be one of the main issues. If it s not possible to create attention, then it is not possible to start an interpretation, decisionmaking and acting process. Obviously, there will still be some kind of interpretation, decisionmaking and acting, and the assembly personnel will continue assembly engines, but the risk of assembly errors increases if one fails to trigger the personnel s attention [2] at the right time and to right data source. Structured translation of data into action [3] to reach a specific goal must be the main focus for the assembly personnel. This requires that the information is available at the right time in the right place and that the assembly personnel have identified a need for the specific information. 2. THE ROLE OF INFORMATION TRIGGERS Fulfill Goal Receiver Knowledge Demand Knowledge Action Transformation/ Interpretation Knowledge Action Demand Transformation/ Interpretation Information Transformation Data Action Transformation/ Interpretation Knowledge Demand Achieve Goal/ Create a Reject Figure 1: Illustration of a goal driven information flow. The purpose of any rational action should be to achieve a goal [3, 4, 5]. This should normally create a demand for information, see figure 1. This demand is triggered by different object states in the work environment, e.g. sounds on and off, lights on and off etc. but it might also be triggered by organisational demands such as quality demands.
In principle, there are four situations that can occur regarding information need versus demand: 1. There is a need but no demand. In this situation, an error will sooner or later occur. This situation can be remedied by introducing appropriate triggers that can create the required demand. 2. There is a need and a demand. This situation is the preferred one. In this situation, the error risk due to a lack of information is lowest. However, this situation still requires that the information available matches the need. Furthermore, if there is a mismatch between need and demand, then a residual state (1) or state (3) will eventually emerge. 3. There is a demand but no need. This situation can be frustrating for the personnel. They have identified a need and have a demand, but the context of e.g. the assembly station doesn t provide them with the (subjectively) needed information. This obviously is a potential future error source; frustration can result in the absence of AIS behaviour in the future, even when an information need is identified. 4. There is no need and no demand. This situation is more or less trivial. 3. THE CENTRAL ROLE OF INFORMATION 3.1 Models Relating Action to Information The OODA loop, see figure 2, (Observe, Orient, Decide, Act) [6] is an example of a model relating the information process to actions. The PDCA wheel is another example. It is often used in quality improvement processes; within industry for instance as a support in TPM. In production engineering, one of the prevailing models is Deming's PDCA cycle. However, since we focus on assembly as a cognitive task, the OODA loop as used in the defence sector may provide a more suitable framework. Observe Orient Act Decide Figure 2: PDCA cycle (some times referred to as Deming s Wheel ) [7] and The OODA-loop. (Boyd s control loop) However, some researchers [8] have suggested that OODA should be redefined to Observe, Interpret, Decide and Act, see figure 3. This would create a more generic model more suitable for manufacturing purposes, and in this case make it possible to map Attention, Interpreting, Decision making and Acting to the OODA model in a better way.
Figure 3: The redefined OODA loop. [8] Like many other models regarding the information process, both the PDCA model and the OODA model assume that the process starts with an identified information demand. This might not be a problem if the information system developer realises that. Within the industry, and not only there, there are signs that there is a lack of understanding regarding how a information search process starts and how it lives on, and in particular how it can live on with a minimum of mental effort. If the information systems design to transfer data and information doesn t support an information search process, then errors may occur due to a lack of information. 3.2 Attention Before it is possible to receive information, one must perceive its presence. The paper Automation and Situation Awareness [9] stresses a very important issue regarding information and information receivers. That is that there is a problem connected to e.g. production systems as well as a pilot's work environment, i.e. the cockpit, and this is active and passive attention, regarding the elements /context in a work environment. Together with active and passive information seeking, passive and active attention is believed to be the main contributors to the problems in the specific work environment [1, 4, 10]. In short, active and passive attention can be explained as: 1. Active attention. Is connected to actively processing information. E.g. driving in intense city traffic where one actively seeks and processes information 2. Passive attention. Driving on an empty motorway, we may register data/information in our surroundings without processing it actively until something unusual happens, e.g. the sight of a speed camera. According to Endsley R. Mica [2] passive attention is a state that can affect the ability to correct, integrate or comprehend information. One interesting part in her discussion is that in a study regarding an automobile navigation task she has found that perceiving information/data correctly was affected negatively if a receiver was in a passive attention state. The receiver was still aware of presented data, still capable of perceiving data, but was less effective when it came to comprehending the meaning of data (i.e. to process data into information) and to relating it to the goal of a task or operation. There are similarities between this study and the work environment at the specific plant where assembly personnel seems to be aware of some data, and perform part of a task correct, but at the same time miss other important issues connected to the same task, although all data/information is available in the context of the work station. In other words, attention plays an important role in how we observe data.
The OODA loop is iterative and nonlinear, and in the iterative process the observation phase is influenced by previous action. 3.3 Interpreting In this case interpreting refers to the process of transforming [11] data into information. To be able to discuss information use, one must first define data and information. One way to describe the difference between data and information is: Data can be described as: 1. A set of symbols in which the individual symbols have a potential for meaning but may not be meaningful to a given recipient. or 2. A set of symbols in which the individual symbols are known, but the combination is meaningless: the semiotics are known; the syntactics are not. or 3. Understandable symbols rejected by the recipient as being of no interest or value, typically because redundant or disbelieved. [11] These descriptions are not mutually exclusive, for instance situation 1) and 3) can occur at the same time. Information can be described as: 1. A message that exists but that is not necessarily sent to, or received by, a given recipient, such as books, unread, in a library, yet deemed significant by someone. In many minds there is no difference between data and this meaning of information. or 2. A message sent to a destination or received by a destination, but not evaluated or understood. The distinction between this and the first definition is small. This definition implies that the message is in some way called to a user's attention, but not assimilated by that person. Information as thing is included herein. or 3. A message understood by the recipient and which changes that person's knowledge base. or 4. Information is an output of the process of converting received messages, data, signs, or signals into knowledge. [11] With this definition as a foundation it is possible to draw a conclusion that there is a connection between interpretations, data and information. The definition of information [11] states that it is an output of a process. This output among other things, such as knowledge, is a base for decision-making and acting. From this it is possible to state that information in the assembly plant context could be: A message that when received, read, interpreted by a recipient create knowledge/change the receiver s knowledge base so that an action can be committed by the receiver that is predicted by the sender. 3.4 Decision-making and Acting Although the subject of linking acting to decision making is an interesting topic in its own right, within the context of this paper these two topics (the second part in OODA) can be discussed as one. Decision-making can be seen from a process perspective as a many-to-one mapping of information to responses. [12] In the specific environment some of the data/information sources are viewed in figure 4.
Figure 4: A simplified visualization of data/information sources in the specific environment. If one considers this visualization together with the mapping statement it is easy to see that there could be different data/information sources that provide assembly personnel with data/information. This of course can directly affect the decision-making process in a negative way. In a developed version of the OODA-loop, see figure 5, it is possible to relate to this as "actors of observation". Figure 5: Developed OODA-loop [13] One step in the decision-making process is change/update of knowledge-base. Knowledge is one of the fundamental information providers within the process. This will not be discussed
in detail in this paper, but it is worth mentioning. According to some researchers [11] one difference between data or even noise and information is the impact on a receiver s knowledgebase. 4 The importance of triggers As mentioned earlier in this paper, a trigger is the signal that creates a change of state regarding attention, from passive to active, and preferable, from active to passive. The hypothesis is that on the shop floor much of the attention is focused on the assembly task, and not on gathering data/information. This selective attention [14] is a part of the human nature, so an information system should provide a possibility to focus the assembler s attention: 1. To the Right place: Where within the workplace context is the data presented? There is evidence that at the specific plant, the personnel search and use data/information from non-reliable sources. 2. At the right time: The trigger and the data must be synchronised so that the need and the demand coexist. At the same time, it is important to note that: 1. The information system should support the personnel so that a change of state, from passive to active is possible, and preferable from active to passive attention is possible. 2. According to some researchers, Kinsbourne in Reason s Human Error [15], it is the nature of the attention process to generate its own extinction. This indicates that a trigger must also contribute to the resurrection of the information demand. 3. It is important that a trigger is used in a way that really supports the personnel. There is evidence at plant that misuse of quality support system have, instead of supporting the personnel, created a feeling of irritation, and the benefit of the system is gone. It is important that triggers are used in a way that is understandable, that using it has a purpose and that the feedback from using it proves this. 5. CONCLUSIONS AND OUTLOOK 5.1 Conclusions In the assembly environment there is evidence, internal rejects, that the personnel don t use the information system in the most effective way. Studies made on the shop floor have viewed that the support system, from a graphical point-of-view are well, not perfect, design, but the identified user do not use them in a way that was anticipated. Input from ongoing projects indicates that one the main reasons for this is the attention levels among the assembly personnel. It is the nature of the attention process to generate its own extinction. If this is the case, and there are indications that support this, one main issues for a IT/IS-system in an manual assembly environment should be support an information search behaviour that is influenced by the human behaviour regarding passive and active attention, e.g. the IT/IS-system must resurrect the attention to a level that create a demand for information. It is believed by the authors that as long as information search can be seen upon as a secondary task, after the main task, assembly, a support for passive and active attention must exist, and to be able to change between the different stats, triggers must be used. 5.2 Outlook The main focus regarding future work will be on triggers, triggers that can change the attention levels, from passive to active and from active to passive, and study the effects that this has on quality. As a foundation for this work the Cognitive Walkthrough method [4] will be
used. This method is used e.g. by software developers to evaluate user task/-s. There is an interest in finding an evaluation method that can be used in an industrial environment with the demands this environment puts on an evaluation method. The result from this evaluation will be used for finding assembly task or part of, where a support by triggers should make an impact on the quality. REFERENCES [1] Bäckstrand, G., De Vin, L., Högberg, D., Case, K. (2005) Parameters affecting quality in manual assembly of engines Proceedings of the 22nd International Manufacturing Conference: Challenges Facing Manufacturing. 31st August to 2nd September 2005. [2] Endsley R. Mica (1999) Situation Awareness and Human Error: Designing to Support Human Performance. Proceedings of the High Consequence Systems Surety Conference, Albuquerque. [3] Johnstone, D; Bonner, M; Tate, M. (2004) "Bringing human information behaviour into information systems research: an application of systems modelling" Information Research, 9(4) paper 191. http://informationr.net/ir/9-4/paper191.html Link available 2005-02-22. [4] Wharton Cathleen, Lewis Clayton (1994). The Role of Psychological Theory in Usability Inspection Methods. In Nielsen Jakob, Mack L. Robert (Eds.) (1994) Usability Inspection Methods. John Wiley & Sons, Inc. New York. [5] Wilson, T.D. (1999) "Models in information behaviour research" Journal of Documentation, 55(3) 249-270 [Available at http://informationr.net/tdw/publ/papers/1999jdoc.html] [6] Blasch, E and S. Plano. (2002). JDL Level 5 fusion model: user refinement issues and applications in group tracking, SPIE Vol. 4729, Aerosense, 2002, pp. 270 279. [7] http://www.hci.com.au/hcisite2/toolkit/pdcacycl.htm. Link available 2006-05-17 [8] De Vin J. Leo, Ng H.C. Amos, Andler F. Sten, Oscarsson Jan. (2005) Information Fusion for Simulation Based Decision Support in Manufacturing FAIM2005, Bilbao, Spain [9] Endsley R. Mica (1996) Automation and Situation Awareness. In R. Parasuraman & M. Mouloua (Eds.) (1996) Automation and Human Performance: Theory and Applications. [10] Machlup Fritz (1980) Knowledge: Its Creation, Distribution and Economic Significance, Volume 1: Knowledge and Knowledge Production Prinston University Press. Princeton, New Jersey. [11] Meadow T. Charles, Yuan Weijing. (1997). Measuring the Impact of Information: Defining the Concepts. Information Processing and Management, Vol. 33. No.6. pp. 697-714 [12] Wickens, C.D.(1999) Engineering Psychology and Human Performance. 3rd. Ed. Prentice-Hall Inc. [13] http://en.wikipedia.org/wiki/ooda_loop. Link available 2006-05-16 [14] Norman A. Donald (1990) The Design of Everyday Things. The MIT Press, London, England. [15] Reason James. (1990) Human Error Cambridge University Press.