Requirements Elicitation Techniques Early

INTRODUCTION

The problem is investigated in the often large scale and complex healthcare systems experiencing uncertainty in knowledge. Apart from the SAMS project, other expert systems that are highlighted in the study as requirements elicitation (RE) techniques include awareness requirements (AwReqs), soft systems methodologies, and joint applications development. Also, along with the detection of the early signs of dementia, The SAMS implied the requirements for the monitoring and interpretation of computer logs concerning the interaction between humans and computers and connecting the computer log analysis to the human behavioral clinical models that are likely to determine the initial signs of dementia. Upon noticing the changes in the behavior of human-computer interaction, the SAMS system alerts the users, notifying them to begin follow-up steps with their doctors [3].

The report then focuses on the subsequent development of problems regarding requirements analysis using the SAMS Project. This is done by reporting the tracks followed in requirements elicitation and the reflections on requirements discovery in medical databases and the bringing about of the causal diagnostic models from the experts. The reporting of the requirements analysis experience relied on monitoring the health of end-users using the SAMS project. The target for requirements discovery experience was by use of a causal model for the prior symptoms of dementia which will be identified by the errors and changes in user behavior evident through the logs of their computer activity. The Delphi-style process which consists of workshops with experts was used together with questionnaires along with the description of the elicitation process [2].

SUMMARY

This report focuses on the requirements elicitation of domain-based knowledge from experts in the cognitive and clinical neurosciences. This is particularly due to uncertain knowledge regarding the complex and large scale healthcare systems which pose significant problems for the requirements engineering. The success of different approaches such as the participatory design or co-designing with relative domain experts has proven to be very sparse. The paper reports the experience of adapting techniques for requirements analysis to address healthcare system problems. The report majorly relies on the Software Architecture for Mental Health Self-Management (SAMS) project to analyze the recordings of email texts, computer activity, to detect the early signs of dementia and the behavioral changes which may indicate the beginning of pathology procedures [1].

Two preliminary approaches are suggested for the requirements analysis. First, knowledge is elicited from the clinical domain experts who specify the diagnostic behaviors of dementia. This knowledge serves to imply the requirements as some kind of causal model that experts solicit in cognitive and clinical neuroscience linking the behavioral metrics to signs of the diagnosis of dementia. The diagnostic tests made on paper serve as starting points for the analysis which is likely to embed the same knowledge in an expert system. Second, the interpretation of user behavior is enabled through the logging of mouse actions, emails, and keyboard use. Also, the logging identifies behavioral changes that are likely to point out the initial diagnostic signs of dementia or the progressive decline in cognitive processes. After these approaches, data mining tools can be used to aid in the mode of discovering hypothesis through the suggestion of requirement questions to determine how the trends and patterns discovered can be interpreted in terms of relative medical implications.

The data mining tools can help in the mode of data-driven approach where aberrant patterns might be discovered by text- and data-mining tools in the logs of interaction between a person and the computer which indicate the potential medical problems [4]. A synthesis of two of the approaches appears at the beginning of the SAMS project to present better prospects to a solution. The questions derived and answered in each approach serve to answer the RE perspective of unknown unknowns for two different purposes. First, there is some uncertainty underlying the level of knownness of the knowledge possessed by domain expert medical practitioners. Second, there are several goals which are partially known to be perceived by the requirements analysts such as the need for a causal model for diagnosis and the need for data mining that is directed by hypothesis.

In the field of knowledge engineering and classic requirements, elicitation techniques have been reviewed and researched extensively leading to the conclusion that different techniques like workshops, scenarios, expert conversations and prototypes can be appropriate and their choice depends on being able to match the techniques to the stakeholders and the problem. Many studies are in agreement that domain knowledge that is possessed by knowledge/requirements engineers can help in bridging the communications gap. The elicitation of unknown expert knowledge is responsive to the popular RE techniques in the knowledge framework point of view. Awareness requirements (AwReqs) are the monitoring requirements of the SAMS project software. AwReqs work to investigate the failure or success of alternative requirements. However, the requirements are regarded in terms of feedback of short-term nature with few variables. While in contrast, the perspective for the study is to monitor complex phenomena (like people) in the long term to point out the deviations in the normal behaviour that is considered healthy and, therefore, requirements to take the remediating steps such as treatments.

The ultimate goal for AwReqs in the SAMS perspective is to observe the state of the health of a user in loops of feedback to encourage actions for follow-ups whenever there is the detection of deviation from patterns which are regarded as normal. There is the tendency of AwReqs to focus on the monitoring of system variables of a small number followed by the adaptation of behaviour with changing parameters and rules in algorithms [5].

ALGORITHMS AND SYSTEMS

Several basic RE techniques, such as observation, interviews, workshops, prototypes, models, and protocols, are described in RE books [6]. This makes the requirements elicitation a mature part of RE. The selection of a model for elicitation technique relies on the matching of the techniques to the RE conditions which consist maturity of requirements, the domain, organizations, and stakeholders. No advantage is found to exist over semi-structured interviews in other techniques, and neither did the use of prototypes and models, which are representations, offer any help. After comparing the elicitation techniques in soft methodology systems, participatory design, and joint applications development (JAD) against the framework of the roles of user designs, communication techniques and activities like prototyping, cognitive (protocols), interviews, group workshops, contextual (ethnography) and representations which are model-driven, Couglan and Macredie concluded that the dialogue-based and collaborative methods that were inclusive of workshops were most effective.

The codification of domain expert knowledge as the rules in knowledge-based systems represents the beginning attempts of the requirement elicitation problems. It was difficult to synthesize the rules from the domain knowledge provided, and the elicitation of the knowledge was less problematic, and generally had many challenges to acquiring knowledge effectively. Several techniques which are successfully used in the elicitation of expert requirement knowledge in different expert system contexts are represented by Kelly’s Personal Contract Theory (PCT) [7]. The PCT holds the notion that the world is represented by people through the element and constructs frameworks that determine how they construe or perceive the world. A construct is regarded as a range of values in poles, and the experts assign values between the construct poles to place a description on the way they perceive an associated element. For instance, a cognitive psychologist can regard executive function as an element (value) within the specialist domain construct. Thus, they may construe the executive dysfunction as a value lying between poles, which represent the diagnostic significance spectrum within the construct that ranges from strongly indicative of dementia, or not indicative of dementia [8].

Efforts to synthesize the utilization of PCT leads to a variety of broadly applied techniques. The most important among the techniques id repertory grid that is representative of the constructs and elements in a matrix. Analysts use card sorting as a way of complementing the repertory grid as a route to understanding the elicited knowledge in a much better way by requesting experts to order and classify concepts in their world against a set of criteria determined by analysts. In the SAMS-like system, card sorting can be of help to an analyst in prioritizing the tasks where data collection and monitoring are done routinely [9]. Although there may exist different variations of card sorting, it is still considered a way that is generally easy for stakeholder engagement and for the successful elicitation of experts in various domains.

Card sorting and repertory grid are the expert knowledge elicitation techniques through which experts can articulate their world using personalized terminologies and provide a way for the domain-ignorant analyst to probe an expert as well as minimizing the problems projection regarding the perceptions of analysts. In domains where there is incomplete and dispersed expert knowledge, the imperfect consensus is commonly found among the expert community members. Card sorting is useful in requirements elicitation from groups, while the Delphi method is used in requirement elicitation from expert groups [9]. The Delphi method was first used in the elicit prediction of the state of the world in the future. Ever since, the Delphi method is used in various problem domains, which include the RE.

Awareness requirements (AwReqs), which are an evolution of the original notion of computational reflection which was developed by Smith, adapt the idea that a system software ought to be subject to requirement introspection [10]. This means that, during run-time, the AwReqs need to be aware of the extremity at which requirements are being satisfied and to support the adaptation courses of action which take advantage of the self-adaptive machinery systems. Therefore, they pose as a method of detecting the failure or success of the requirements and this can be in terms of quality constraints, tasks, domain assumptions and goals. The requirements are then mapped to loops of feedback which have monitoring capabilities which can determine the levels of goal satisfaction.

The approach for an expert conversation workshop yielded good results by following the process of knowledge acquisition even though the experience illustrated that the elicitation of requirements based on knowledge remains difficult a task regardless of practice that was done for decades for its refinement [12]. Some of the challenges which were faced despite the appropriate amount of knowledge acquired by the RE team included knowledge and language gap. The experience suggests that in order to bridge the gap of communication in the elicitation of knowledge with experts requires domain knowledge of considerable nature from requirements engineers, which is in contrast to the conventional requirements to the acquisition of domains. The extensive modification of the Delphi method throughout the process served as an indicator of the advisable adaptive and flexible elicitation approaches [13]. The goal reflection of the study appeared too ambitious and improbable because there lacked any mature model of diagnosis of the behavioral and cognitive signs of the existence of the beginning of dementia.

RESULTS AND DISCUSSION

The SAMS project acquires contribution from the RE experience to the extent that the concept of awareness requirement in domains which are complex, and where the monitoring object is not the system design but an entity in the domain, like an organization, the environment, and people. The second contribution surpasses the framework of the unknowns for the RE to explore the precincts of knowledge that is known with development teams in the domain [11]. A new experimental approach to the RE in complex domain is advocated for. Also, a fundamental approach for knowledge acquisition is evaluated based on the Delphi techniques for the RE.

It was concluded to evaluate the maturity level on knowledge concerning domains and the domain is uncertain, the elicitation exercise would be framed according to a road map of research. The exercises would include the discovery of the unknowns, at the point where more research was needed, and the unknowns, which have a confidence level of less than 100%, and the mature knowns. The experience shows the level of knowness of RE at project initiation. Partial knowledge is evident in the fact that a causal diagnostic model and expert knowledge were required, even though the domain knowledge that existed was “unknown.”It transpired that this knowledge was largely unknown by the experts. In addition, the requirements which are unknown by the experts and analysts came up during the project and they included the individual problem of variation, which was the “unknown unknown” from the experts’ and analysts’ point of view.

Expert recruitment served as another problem of the known kind which was encountered even after selecting an array of experts on research and the clinical practices regarding dementia. Crowd source requirements is an approach that has gained considerable interest in the RE and would be useful to escape the recruitment limitations. It is expected that crowd-sourcing does not turn into a panacea for recruitment of experts even though it presents a possibility for future research. Problem-framing is important and the study had different processes adopted in the workshops in a rapid manner. It is also difficult to understand problem framing using the internet. A considerable quantity of material for briefing was required for the promotion of expert comprehension of the requirements and online motivation to digest this information is likely to be minimal. Also, the motivating and selecting the right kind of experts, even without the aid of crowd-sourcing panels of selection, was probably a hard task.

The limitation of the number of experts was identified in the study and there remains the belief that repeating the study with an added number of experts will not yield any results which are radically improved. The domain complexity and the tasks on the precincts of expert knowledge, consensus improvement would be a hard task. Looking back at the approach used in the study, it may have been relevant in the discovery of exploratory knowledge in a domain that is particularly known and most techniques for knowledge elicitation assume domains with stable and mature knowledge.

The study also reveals that a range of approaches to RE are adopted in healthcare, and this range is from JAD workshops and interviews, machine learning and formal modeling for analyzing dependencies which are traceable in the healthcare documents, to the ethnographic work practice investigation. The diverse approaches are a reflection of the broad range of applications within the domain of healthcare. The interactive approach appears successful in requirements which relate to the user interface, prototypes, and scenarios even though the hidden requirements which are present in complex domains may be difficult to elicit.

REFERENCES

Sutcliffe AG, Rayson P, Bull C, Sawyer P (2014) Discovering affect-laden requirements to achieve system acceptance In: Proceedings 22nd IEEE International Requirements Engineering Conference (RE’14). IEEE Computer Society Press, Los Alamitos, pp 173-182.
Brown BB (1968) Delphi process: a methodology used for the elicitation of opinions of experts. (No. RAND-P-3925). Rand Corporation, Santa Monica CA
Mioshi E, Dawson K, Mitchell J, Arnold R, Hodges JR (2006) The Addenbrooke’s Cognitive Examination Revised (ACE-R): a brief cognitive test battery for dementia screening. Int J Geriatr Psychiatry 21:1078–1085.
Han J, Kamber M, Pei J (2011) Data mining: concepts and techniques. Elsevier, Amsterdam.
Souza VES, Lapouchnian A, Robinson WN, Mylopoulos J (2013) Awareness requirements. In: de Lemos R, Giese H, Muller HA, Shaw M (eds) software engineering for self-adaptive systems II. Springer, Berlin, pp 133–161.
Gause DC, Weinberg GM (1989) Exploring requirements: quality before design. Dorset House, New York.
Kelly GA (1977) Personal construct theory and the psychotherapeutic interview. Cognit Ther Res 1(4):355–362.
Tan FB, Hunter MG (2002) The repertory grid technique: a method for the study of cognition in information systems. MIS Q 26:39–57.
Paul CL (2008) A modified Delphi approach to a new card sorting methodology. J Usability Stud 4(1):7–30.
Smith B, (1982) Reflection and semantics in a procedural programming language. Ph.D. thesis, MIT Laboratory for Computer Science Report MIT-TR-272. MIT, Cambridge.
Linstone K, Turoff M (1975) The Delphi method: techniques and applications. Addison Wesley, Reading.
Davis A, Dieste O, Hickey A, Juristo N, Moreno AM (2006) Effectiveness of requirements elicitation techniques: empirical results derived from a systematic review. In: Proceedings 14th IEEE international symposium on requirements engineering (RE’06). IEEE Computer Society Press, Los Alamitos, pp 176–185.
Green B, Jones M, Hughes D, Williams A (1999) Applying the Delphi technique in a study of GP’s information requirements. Health Soc Care Community 7(3):198–205