Intelligent Agile Method Framework

Marko Jankovi

, Marko Bajec

, Ghazaleh Khodabandelou

, Rebecca Deneckere

, Charlotte Hug

and Camille Salinesi

Faculty of Computer and Information Science, University of Ljubljana, Tr

ska cesta 25, SI-1000 Ljubljana, Slovenia

Centre de Recherche en Informatique, University of Paris 1 Panth

eon-Sorbonne, Paris, France

Keywords:

Situational Method Engineering, Software Development Improvement, Process Mining.

Abstract:

The paper addresses the problem of the low usage of software development methods in software development

practice. This has been recognized as one of the key reasons for failures in software development projects

and a contributor to the low quality of software. We introduce a novel approach that could help to improve

the maturity of software development processes. The approach is based on the method engineering principles

taking into account the limitations that hinder its use in practice. The main objective of our research is to show

that the method engineering concepts are applicable in real settings and that could contribute to the higher

quality of software development processes and their products.

1 INTRODUCTION

According to the literature, the use of software devel-

opment methods (SDM) in software industry proves

beneﬁcial as it contributes to the higher quality of the

process and its product, i.e. the developed software

(van de Weerd et al., 2006), (Bekkers et al., 2008),

(Bajec et al., 2007), (Karlsson and

Agerfalk, 2004),

(Fitzgerald and Hartnett, 2005). Despite these evi-

dent beneﬁts, the studies on the maturity of the soft-

ware development discipline show that a large percent

of software development companies do not have their

SDMs documented and those that have, do not really

follow them or do not follow them rigorously when

developing software. Several reasons have been iden-

tiﬁed in the past to explain this behavior. Two of them

seem to be most important: (1) rigidity, which does

not allow SDMs to be adapted to speciﬁcs of a par-

ticular project, and (2) their social inappropriateness

– i.e. prescribed methods are not based on the knowl-

edge, culture and attitude of the developers and other

stakeholders of the project (Riemenschneider et al.,

2002), (Mohan and Ahlemann, 2011), (Mirbel and

Ralyt

e, 2006), (Karlsson and

Agerfalk, 2004).

The above phenomenon has been widely acknowl-

edged in the research community and to some extent

also among the practitioners. The related problems

are severe. According to the research conducted by

McKinsey in collaboration with the University of Ox-

ford (Bloch et al., 2012), half of all large IT projects

(projects with initial price exceeding $15 million) run

over budget, over time or deliver less value than pre-

dicted. Furthermore, study conducted by Standish

Group (www.standishgroup.com), compiled into the

CHAOS Report 2009, reveal that only 16% of ob-

served projects have been ﬁnished in time; 32% were

terminated before they were completed; and 52%

were completed behind their schedule and involved

higher cost. All these facts show that this is not

an insigniﬁcant problem and that there is a need for

new ideas and approaches to confront with these chal-

lenges.

In this paper, we describe an approach that could

help to improve the maturity of software development

processes by circumventing the problems that hinder

the use of SDMs in practice. On the long run our

rather ambitious goal is to lower the risk for IT project

failures and to improve the overall quality of the de-

veloped software.

The approach is based on the method engineering

principles and represents a continuation of our past re-

search in this ﬁeld (Bajec et al., 2007), (Vavpotic and

Bajec, 2009), (

Zvanut and Bajec, 2010), (Hug et al.,

2012), (Deneckere et al., 2008).

In the paper we outline the main idea (Section 3),

provide its background (Section 2), describe the re-

search method (Section 4), and ﬁnally explain the re-

maining steps to be done (Section 5).

232

Jankovi

c M., Bajec M., Khodabandelou G., Deneckere R., Hug C. and Salinesi C..

Intelligent Agile Method Framework.

DOI: 10.5220/0004562702320237

In Proceedings of the 8th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE-2013), pages 232-237

ISBN: 978-989-8565-62-4

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

2 BACKGROUND

For decades, researchers have been striving to develop

new approaches and solutions as a response to low

usage of SDMs in practice. We can see the emer-

gence of conﬁgurable methods and supporting tools

(e.g. IBM Rational Method Composer (RMC), Mi-

crosoft Solutions Framework) that support conﬁgura-

tion and tailoring of SDMs to better suite the charac-

teristics of a particular project and organization (Gar-

cia et al., 2011). However, although these tools of-

ten also include best practices, formed on the basis of

vast experiences in software development, they still

require speciﬁc knowledge in order to be used efﬁ-

ciently. To conﬁgure a method using IBM Rational

Method Composer, for example, we need to know

RUP in great details.

Another related domain is light and agile meth-

ods, which are more user-oriented, less complex and

thus easier to implement and use (Laanti et al., 2011),

(Dyb

a and Dingsøyr, 2008). In general, agile and

light methods are more adoptable, as they strive to be

closer to developers and do not bring such overhead

as traditional methods. But the fact is that compa-

nies keen of agile and light methods rarely use spe-

ciﬁc methods but rather just follow their vision. This

again leads to the fact that their ways of working are

not documented and that valuable knowledge gained

on development projects remain only in the heads of

individual developers.

Very related to our work is the discipline called

Method engineering (ME). Brinkkemper (1996) has

deﬁned ME as an engineering discipline to design,

construct and adapt methods, techniques and tools for

the development of information system. A speciﬁc

sub-discipline of ME is situational method engineer-

ing (SME), which tends to support creation of meth-

ods from method parts or tailoring existing methods

to meet requirements of a speciﬁc project. A number

of approaches of SME (Ralyt

e and Rolland, 2001),

(Brinkkemper et al., 1998), (Deneckere et al., 2008),

have been proposed and developed, but despite the

fact that these methods are designed and elaborated

to be adapted, empirical researches show that their

application in practice is rare (Mirbel and Rivieres,

2002), (Ralyt

e and Rolland, 2001). A reason for the

latter can be found in the fact that creating and tai-

loring of methods is time-consuming and requires a

considerable commitment of the developers and other

stakeholders. Furthermore, to apply SME we need

a person that is familiar with numerous methods and

thus capable of constructing new methods by compos-

ing different fragments together. Furthermore, SME

approaches often do not consider social and techni-

cal aspects. This leads to prescribed methods that are

socially or technically noncompliant with the organi-

zation and developers, and consequently not accepted.

As an answer to that issue Bajec et al. (2008) have de-

veloped an innovative approach, called Agile Method

Framework (AMF). Main vision of this approach was

to bring SME principles closer to practitioners, by

considering social and technical aspects that inﬂuence

on how practitioners perceive usefulness of methods

as guidelines for their everyday work. Although the

results showed that AMF was in practice perceived

as useful, it was not fully accepted in the participat-

ing software companies, due to the fact that it still re-

quired the active involvement of developers. A more

detail review of the past research in the ﬁeld of SME

can be found in the work of Henderson-Sellers et al.

(2010) .

3 INTELLIGENT AGILE

METHOD FRAMEWORK

The main idea of the approach that we call Intelli-

gent Agile Method Framework (in further text iAMF)

is to provide methods and tools to (a) help compa-

nies capture the software processes they are de-facto

using (i.e. the real ways of working when develop-

ing software), (b) help them follow these processes

consistently and ﬁnally (c) help them to continuously

improve and document what they do on IT projects.

While there are other approaches striving to the same

goal, the peculiarity of our approach is that it will

require only insigniﬁcant effort from the developers

and other stakeholders of the project. The latter is

one of the main reasons why similar initiatives and

approaches have not succeeded to penetrate in real

practice (Mirbel and Ralyt

e, 2006), (Karlsson and

Agerfalk, 2012).

The ﬁrst step in iAMF is to capture how projects

are typically performed in a particular company. Our

hypothesis is that if prescribed methods are not just

some off-the-shelf methods (such as RUP, SSADM

etc.) but rather based on the knowledge and experi-

ences of the company employees, and take into ac-

count the real ways of working in the company, then

it is much more likely that developers will follow

them rigorously. Instead of involving developers to

describe how they do projects, iAMF will learn from

the observation on what developers and other stake-

holders actually do on the projects, how they per-

form and act, and based on that create (document)

base methods. With a base method we denote a semi-

formal description of how a certain company is per-

forming its projects. Note that a particular company

IntelligentAgileMethodFramework

233

Figure 1: High-level architecture of iAMF supporting toolset.

might employ different software processes for differ-

ent projects. In this sense, the base method repre-

sents an overall method that comprises all the differ-

ent processes that a company is using when perform-

ing its projects. With the ﬁrst step of the iAMF the

respected software process maturity rises to level 3

(documented).

Once base methods are captured, iAMF can be

used to supervise project performance and guide de-

velopers reusing the knowledge and past experience.

Any declination from base methods would be de-

tected and either prevented or treated as a new knowl-

edge and used to supplement the base method. By

this, iAMF facilitates continuous process improve-

ment, which helps to raise the maturity of the process

to level 4 (managed) or even 5 (optimized).

3.1 iAMF Supporting Toolset

In this section, a high-level architecture of the iAMF

toolset is presented (Figure 1). The purpose of the

toolset is to facilitate the iAMF and will be used in

the evaluation phase (see the section on Research Ap-

proach) to validate our approach. Further, we brieﬂy

describe the toolset main components, their purposes

and communication.

3.1.1 Documenting Base Methods and

Metamodels

To support the enactment of software process follow-

ing some prescribed SDM, the ﬁrst step is to capture

and document the prescribed method, i.e. the base

method. Even though it is generally accepted that

elaboration of SDM have a positive inﬂuence on the

software development process (e.g. for the sake of

knowledge reuse and quality assurance), companies

very much differ in the level of details they believe is

useful to capture and document (Garc

ıa et al., 2011).

Furthermore, keeping ways of working documented

and up-to-date is “expensive” and time-consuming.

One of the goals of the iAMF toolset is to enable ef-

ﬁcient and user friendly capturing of base methods

and metamodels that underpin them. Companies can

select from metamodels that support known methods,

such as RUP, and adapt them to suit their purposes and

speciﬁcs, or to create new metamodels from scratch

(Bajec and Vavpotic, 2008). In the later phases meta-

models are crucial to check completeness and consis-

tency of the documented base methods. The toolset

also supports the creation of the map intentional pro-

cess models for intention mining purpose (see Inten-

tion mining). The described functionality is avail-

able through graphical user interface, which allows

method engineers to construct or update base methods

and metamodels, simply by dragging and dropping el-

ENASE2013-8thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

234

ements on the screen and connecting them to express

their relationships. For each metaelement, method en-

gineers can enter detailed information, including tem-

plates that are then used to guide the documentation

of their instances.

3.1.2 Intention Mining

Intention Mining aims to infer user intentions from

event logs (Khodabandelou et al., 2013b). Intention

mining uses event logs as input and allows (1) iden-

tifying intentional process models, (2) checking the

conformity between a prescribed intentional model

and its enactment, (3) identifying the gaps between

the model and the traces and (4) providing recommen-

dations to users at run-time, based on their reasoning

behind their activities (Khodabandelou et al., 2013a).

Intention mining is an important part of the iAMF

toolset. Its purpose is to gather information on inten-

tions the developers were following when perform-

ing various tasks and activities on past projects. For

this purpose trace-based tools will be employed. In-

stead of developing a completely new tool, we con-

ducted a research (Khodabandelou et al., 2013c),

which showed that open source tool Snare could be

efﬁciently used for our purposes, as it covered almost

all our requirements. The Snare will be used to col-

lect traces of activities and retrieve information about

what and how developers perform during projects.

Our expectation is that by the analysis of traces we

will be able to reconstruct intentions that guided de-

velopers when performing speciﬁc tasks and based

on that create intentional process models for further

guidance.

3.1.3 Providing Methodological Guidance

Once a base method and metamodels are captured,

the guidance can be provided to developers who are

novice or not familiar enough with the base method

their company is prescribing. To improve and person-

alize methodological guidance, collected traces will

be used to automatically categorize users according

to their proﬁles. Based on that and together with the

context at hand and history records, the users will be

provided with an advice on what step to take next in

a particular situation. Furthermore, process models

combined with examples of best practices will guide

developers to efﬁciently perform their activities. In

order to give users good overview and efﬁcient guid-

ance, the iAMF toolset will provide graphical inter-

face for users to browse between various base meth-

ods, to get detailed information on a particular ele-

ment, or just to scroll through the base method visu-

alized as a process diagram or a navigation tree.

3.1.4 Controlling the Use of Base Method

The most important feature of iAMF is the ability to

observe the method in action and to identify any de-

viation from the method the team members are ex-

pected to follow (base method). To detect deviations,

the proposed framework relies on two sources of in-

formation: (1) the base method that prescribes how

different kinds of projects should be carried out (see

also the section Documenting Base Methods) and (2)

the real method or method in action, which is not doc-

umented but can be inferred by observing what devel-

opers do and how they perform on projects.

To infer the in-action method we will intercept ev-

ery action a user does on the Revision control system

(RCS) (i.e. checking an artifact in or out), and on the

basis of (1) information gained from collected traces,

which hold information about the activities performed

before the action, (2) information extracted from the

content of the artifact with text mining, and (3) infor-

mation mined from RCS logs, predict current activity

and link it with the base method. If the action will

not be in accordance with the base method, the devel-

oper performing the action will be asked to provide

the explanation for the deviation. Sometimes the de-

viation will be unintentional and the developer will

be instructed to follow the prescribed method while

in other cases the deviation will be a result of a spe-

ciﬁc situation not seen before and will be thus used

to supplement the base method. For the beginning,

only support for Subversion and GIT that are major

revision control systems in use today, will be imple-

mented. However, a generic interface will be pro-

vided to support the information capture from a va-

riety of application and systems developers are using

to facilitate their work in software development.

With all the collected data we expect to be able to

infer the “path” that has been taken through the base

method as well the current stage of the project accord-

ing to the base method (e.g. which activities and arti-

facts have already been performed/produced and what

steps remains to be done). The collected data will also

be used to provide guidance for the developers on the

project.

3.1.5 Supplementing the Base Method

The deviations from base methods might have differ-

ent reasons. It might happen, for example, that the

user made a mistake, e.g. by taking some action that

was not expected according to the base method – i.e.

he checked in an artifact A, which was not yet re-

quired, while some other artifact B, which would need

to be created before the A, was not yet in the RCS.

IntelligentAgileMethodFramework

235

But on the other hand, there might be speciﬁc circum-

stances that leaded the user to react differently than

prescribed by the base method. In the ﬁrst case, the

user is expected to eliminate or compensate the mis-

take, e.g. by performing the activity he forgot and by

creating the missing artifact, while in the latter he has

to supplement the base method so that the new cir-

cumstances including possible new method elements

and relations would be covered for the future cases. In

this way the base method continuously complements

and improves.

4 RESEARCH APPROACH

The research positioned in this paper presents a join

work of researchers from the University of Ljubljana

and University of Paris 1. Both research groups are

experienced in the SME research ﬁeld with good re-

search record. Their joint concern is to improve soft-

ware development practice by the application of the

SME approaches and principles. Having good con-

nections and experience with local industry, the teams

are seeking for cross-cultural collaboration that could

provide deeper understanding of the subject matter

and provide solutions that could be useful beyond

their local environments.

In this section we brieﬂy outline the main phases

of the research:

Phase 1: Initiation. In the initiation phase we will

take care for the research environment. We will se-

lect the RCS systems that are most popular among

software companies so that we could then develop in-

terfaces to capture required information. Similarly, a

workﬂow-based enactment system will be selected to

be able to get detailed information on software pro-

cesses and speciﬁcally on developers’ behavior when

performing tasks within these processes. As a part of

the initiation phase we will try to identify and stimu-

late companies for their participation on the project.

To be able to evaluate the suggested approach in real

settings, we will invite several software companies

from Slovenia and France to participate in the re-

search. To cover the variety of the software industry,

we will try to involve companies of different sizes,

organizations, cultures, experiences, . . .

Phase 2: Implementation. In this phase we will im-

plement the required iAMF tools. These will include

all the components described in section 3, including

algorithms for mining intentions from process logs.

Phase 3: Testing. The iAMF will be ﬁrst tested

within the university environment (at both universi-

ties). A group of students will use the framework as

a part of their seminar work. The purpose of these

tests will be to detect and eliminate software bugs, as

well as to get general opinion on the usability of the

framework. iAMF will be then improved based on the

testing results.

Phase 4: Evaluation. In the evaluation, our goal is

to empirically conﬁrm the usability of the approach.

This is not an easy task, as it cannot be performed

within university environment. To get reliable feed-

back, several companies will be invited to evaluate the

framework. With the evaluation we will try to con-

ﬁrm the main hypotheses: (H1) If prescribed meth-

ods were based on real ways of working and up to

date then developers would follow them much more

rigorously. (H2) iAMF can provide efﬁcient support

to project members by helping and guiding them in

software development projects. (H3) Developers and

other users do not perceive iAMF as an additional bur-

den and are willing to use it. (H4) By using iAMF,

companies can raise the maturity of their software

processes to at least level 3.

5 CONCLUSIONS

In this position paper we outlined a research that aims

at providing a novel approach for improving software

development practice. The approach consists of a set

of methods and tools that will help companies to elab-

orate and continuously improve their software pro-

cesses, which will in-turn make them more useful and

usable for developers and other stakeholders on soft-

ware development projects. The research proposed

in this paper is due to its time-consuming evaluation

phase longitudinal. Giving any concrete conclusion at

this moment would be premature and risky.

REFERENCES

Bajec, M., Vavpoti

c, D., and Krisper, M. (2007). Practice-

driven approach for creating project-speciﬁc software

development methods. Information and Software

Technology, 49(4):345–365.

Bajec, M. and Vavpotic, D. (2008). A framework and

tool-support for reengineering software development

methods. Informatica, Lith. Acad. Sci., 19(3):321–

344.

Bekkers, W., van de Weerd, I., Brinkkemper, S., and

Mahieu, A. (2008). The inﬂuence of situational fac-

tors in software product management: An empirical

study. In Second International Workshop on Software

Product Management, 2008. IWSPM ’08, pages 41 –

48.

ENASE2013-8thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

236

Bloch, M., Blumberg, S., and Laartz, J. (2012).

Delivering large-scale it projects on time,

on budget, and on value. Retrieved from

http://www.mckinsey.com/insights/business technology/.

Brinkkemper, S. (1996). Method engineering: engi-

neering of information systems development meth-

ods and tools. Information and Software Technology,

38(4):275–280.

Brinkkemper, S., Saeki, M., and Harmsen, F. (1998). As-

sembly techniques for method engineering. In Pro-

ceedings of the 10th International Conference on Ad-

vanced Information Systems Engineering, CAiSE ’98,

page 381–400, London, UK, UK. Springer-Verlag.

Deneckere, R., Iacovelli, A., Kornyshova, E., and Souveyet,

C. (2008). From method fragments to method ser-

vices. In Proc. Of EMMSAD’08, pages 80–96, Mont-

pellier, France.

Dyb

a, T. and Dingsøyr, T. (2008). Empirical studies of agile

software development: A systematic review. Informa-

tion and Software Technology, 50(9–10):833–859.

Fitzgerald, B. and Hartnett, G. (2005). A study of the use of

agile methods within intel. In Baskerville, R., Math-

iassen, L., Pries-Heje, J., and DeGross, J., editors,

Business Agility and Information Technology Diffu-

sion, volume 180 of IFIP International Federation for

Information Processing, pages 187–202. Springer US.

Garc

ıa, J., Amescua, A., S

anchez, M.-I., and Berm

on,

L. (2011). Design guidelines for software processes

knowledge repository development. Information and

Software Technology, 53(8):834–850.

Garcia, F., Vizcaino, A., and Ebert, C. (2011). Process man-

agement tools. IEEE Software, 28(2):15 –18.

Henderson-Sellers, B. and Ralyt

e, J. (2010). Situational

method engineering: State-of-the-art review. Journal

of Universal Computer Science, 16(3):424–478.

Hug, C., Deneck

ere, R., and Salinesi, C. (2012). Map-tbs:

Map process enactment traces and analysis. In RCIS,

pages 1–6.

Karlsson, F. and

Agerfalk, P. J. (2004). Method conﬁgura-

tion: adapting to situational characteristics while cre-

ating reusable assets. Information and Software Tech-

nology, 46(9):619–633.

Karlsson, F. and

Agerfalk, P. J. (2012). MC sandbox: De-

vising a tool for method-user-centered method con-

ﬁguration. Information and Software Technology,

54(5):501–516.

Khodabandelou, G., Hug, C., Deneck

ere, R., and Salinesi,

C. (2013a). Process mining versus intention mining.

In Procs. EMMSAD’13, Valencia, Spain.

Khodabandelou, G., Hug, C., Deneck

ere, R., and Salinesi,

C. (2013b). Supervised intentional process models

discovery using hidden markov models. In Procs.

RCIS’13, Paris, France.

Khodabandelou, G., Hug, C., Deneck

ere, R., Salinesi, C.,

Bajec, M., Kornyshova, E., and Jankovi

c, M. (2013c).

A systematic review and selection of cots products

to trace method enactment. In Proceedings of the

21st European Conference on information Systems,

Utrecht, Netherland.

Laanti, M., Salo, O., and Abrahamsson, P. (2011). Ag-

ile methods rapidly replacing traditional methods at

nokia: A survey of opinions on agile transformation.

Information and Software Technology, 53(3):276–

290.

Mirbel, I. and Ralyt

e, J. (2006). Situational method en-

gineering: combining assembly-based and roadmap-

driven approaches. Requirements Engineering,

11(1):58–78.

Mirbel, I. and Rivieres, V. d. (2002). Adapting analysis

and design to software context: The JECKO approach.

In Bellahs

ene, Z., Patel, D., and Rolland, C., editors,

Object-Oriented Information Systems, number 2425

in Lecture Notes in Computer Science, pages 223–

228. Springer Berlin Heidelberg.

Mohan, K. and Ahlemann, F. (2011). What methodology

attributes are critical for potential users? understand-

ing the effect of human needs. In Mouratidis, H. and

Rolland, C., editors, Advanced Information Systems

Engineering, volume 6741 of Lecture Notes in Com-

puter Science, pages 314–328. Springer Berlin / Hei-

delberg.

Ralyt

e, J. and Rolland, C. (2001). An assembly process

model for method engineering. In Dittrich, K. R.,

Geppert, A., and Norrie, M. C., editors, Advanced

Information Systems Engineering, number 2068 in

Lecture Notes in Computer Science, pages 267–283.

Springer Berlin Heidelberg.

Riemenschneider, C., Hardgrave, B., and Davis, F. (2002).

Explaining software developer acceptance of method-

ologies: a comparison of ﬁve theoretical mod-

els. IEEE Transactions on Software Engineering,

28(12):1135 – 1145.

van de Weerd, I., Brinkkemper, S., Souer, J., and

Versendaal, J. (2006). A situational implementation

method for web-based content management system-

applications: method engineering and validation in

practice. Software Process: Improvement and Prac-

tice, 11(5):521–538.

Zvanut, B. and Bajec, M. (2010). A tool for IT process

construction. Information and Software Technology,

52(4):397–410.

Vavpotic, D. and Bajec, M. (2009). An approach for concur-

rent evaluation of technical and social aspects of soft-

ware development methodologies. Information and

Software Technology, 51(2):528–545.

IntelligentAgileMethodFramework

237