J. Filipe and J. Cordeiro (Eds.): ICEIS 2010, LNBIP 73, pp. 340–356, 2011.
© Springer-Verlag Berlin Heidelberg 2011
Using Cases, Evidences and Context
to Support Decision Making
Expedito Carlos Lopes1, Vaninha Vieira2, Ana Carolina Salgado3,
and Ulrich Schiel1
1 Federal University of Campina Grande, Computing and Systems Department
P.O. box 10106 Campina Grande, PB, Brazil
2 Federal University of Bahia, Computer Science Department, Salvador, BA, Brazil
3 Federal University of Pernambuco, Informatics Center, Recife, PE, Brazil
Abstract. Evidence-Based Practice (EBP) represents a decision-making process
centered on justifications of relevant information contained scientific research
proof found in the Internet. Context is a type of knowledge that supports identi-
fying what is or is not relevant in a given situation. Therefore, the integration of
evidence and context is still an open issue. Besides, EBP procedures do not
provide mechanisms to retain strategic knowledge from individual solutions,
which could facilitate the learning of decision makers, preserving evidences
used. On the other hand, Case-Based Reasoning (CBR) uses the history of
similar cases and provides mechanisms to retain problem-solving. This paper
proposes the integration of the CBR model with EBP procedures and Context to
support decision making. Our approach includes a conceptual framework ex-
tended to support the development of applications that combines cases, evi-
dence and context, preserving the characteristics of usability and portability
across domains. An implementation in the area of crime prevention illustrates
the usage of our proposal.
Keywords: Case-based reasoning, Evidence-based practice, Context, Decision
making, Conceptual framework.
A problem identified inside an organization is the element that triggers the process of
decision making. When a problem presents difficulties associated with the absence of
available solutions, or even when the problem demands a great effort to understand
why it happens or what are its origins, regardless of its solution, it is considered a
complex problem . In such cases, the search for knowledge beyond the organiza-
tional environment can be necessary.
Evidence-Based Practice (EBP) is usually employed in Medicine with the focus
on providing effective counseling to assist patients with terminal or chronic diseases
on making decisions. It involves complex and conscientious decision-making based
on the best research evidence found in researches over Internet sites [2, 3]. Therefore,
EBP procedures do not provide mechanisms to retain strategic information and
Using Cases, Evidences and Context to Support Decision Making 341
knowledge from individual solutions. This history could facilitate the learning of
different end-users, in the future, preserving used evidences, since they can later be
modified or removed from the Internet.
On the other hand, an important model from the Artificial Intelligence area is Case-
Based Reasoning (CBR). CBR uses the history of similar cases to support decision
making providing mechanisms to retain problem-solving in Case Base .
Since both EBP procedures and CBR techniques are important to support decision-
making, the integration of these two paradigms constitutes an interesting research
topic to support problems solution, especially for complex problems.
Besides, according to Dobrow et al. , the EBP procedures represent a decision-
making process centered on justifications of relevant information. Defining relevant
information is not an easy task. Context is a type of knowledge used to support the
definition of what is or is not relevant in a given situation . Applying EBP proce-
dures to a particular patient case, for example, implies considering different contextual
information regarding the generation of evidences and the patient case itself. Thus,
“the two fundamental components of an evidence-based decision are evidence and
context, and the decision-making context can have an impact on evidence-based deci-
sion-making.” . But, the integration of evidence and context is still an open issue
and, in fact, the evidence retrieval with contextual information can facilitate the reuse
of evidence-based decision-making justifications involving similar situations.
Systems that use context apply it to filter out and share more useful information so
that this information can meet users’ needs. Thus, context is a significant tool to op-
timize a system’s performance and to reduce search results. Filtering mechanisms
avoid more explicit interactions of the user with the application [6, 7].
This paper proposes the integration of Case-Based Reasoning model with Evi-
dence-Based Practice procedures and the usage of context to support filtering mecha-
nisms on decision making when the solution is gathered outside the organizational
environment through research evidence.
Since EBP can be applied to different areas (e.g. Health in general, Education, So-
cial Work, Crime Prevention and Software Engineering in Computer Science), we
extended a conceptual framework, that represents the integration of evidence and
context, to incorporate case structure justified by evidence, preserving the characteris-
tics of usability and portability in domains that use EBP. This framework supports
system designers in the conceptual modeling phase, providing more agility, transpar-
ency and cohesion between models. To illustrate our proposal usage, we implemented
the framework extended in the area of crime prevention.
The rest of the paper is organized as follows. The key concepts regarding the main
themes of this work are described in Section 2. Section 3 presents the conceptual
framework extended using UML and the integration of EBP with context to the CBR
model. Section 4 presents the application of the framework extended in the area of
Crime Prevention. Related works are described in Section 5. Finally, Section 6 pre-
sents our conclusions and directions for further work.
This section defines context and provides an overview of Evidence-Based Practice
and Case-Based Reasoning.
342 E.C. Lopes et al.
There are several definitions of context. A classical definition (highly referred) is
proposed by Dey and Abowd  which states that context is “any information that
characterizes the situation of an entity, where this entity is a person, place or object
considered relevant in the interaction between the user and an application”.
Context can also be seen as a set of information items (e.g. concepts, rules and
propositions) associated with an entity . An item is considered part of a context
only if it is useful to support the resolution of a given problem. This item corresponds
to a contextual element defined as “any data, information or knowledge that enables
one to characterize an entity on a given domain” .
Contextual information regarding acquisitions is: (i) given by the user, whether
from persistent data sources or from profiles; (ii) obtained from a knowledge base;
(iii) obtained by means of deriving mechanisms; or (iv) perceived from the environ-
ment . It is usually identified through the dimensions why, who, what, where, when
and how .
Other important concept is related to the attention focus of the decision maker. One
step in the task execution or problem-solving process is known as focus. The contex-
tual elements should have a relevant relationship to the focus of a human agent (or
software agent). In general, focus is what determines which contextual elements
should be instantiated .
2.2 Evidence-Based Practice
According to Thomas and Pring , in general, information labeled as evidence is
those whose collection had concerns about its validity, credibility and consistency
with other facts or evidence. In relation to its credibility, the authors categorize
evidence in three ways:
1. Based on professional practice, as a clinical or crime examination;
2. Generated by a process involving scientific procedures with a proven history in
producing valid and reliable results, for example a collect performed by biomedical;
3. Based from published research that corresponds to critical reviews of the area,
such as randomized clinical trial.
“Evidence” in EBP, also called “research evidence”, corresponds to the third category
above and means a superior type of scientific research proof, such as generated
through systematic review1 in the highest level. These published researches are avail-
able in reliable data bases, usually found on sites over the Internet, and carried out by
independent research groups . This is the concept of evidence applied in this paper.
Evidence-Based Practice (EBP) involves complex decision-making, based on
available research evidence and also on characteristics of the actor of the problem,
his/her situations and preferences.
1 A systematic review (SR) is a review that presents meticulous research and critical evalua-
tions of primary studies (case study, cohort, case series, etc.), based on research evidence
related to a specific theme. It contains analysis of qualitative results conducted in distinct
locations and at different times. Meta-analysis is a SR of qualitative and quantitative charac-
Using Cases, Evidences and Context to Support Decision Making 343
In Medicine, EBP primary focus is to provide effective counseling to help patients
with terminal or chronic illness to make decision in order to cure the illness, extend or
increase the quality of their life . What is objectively searched is “the integration of
best evidence from research, clinical skill and preferences of the patient, regarding
their individual risks and the benefits of proposed interventions” .
In crime prevention, EBP involves the correlations practice that has been proven
through scientific research, aimed at reducing the recidivism of offenders. EBP pri-
marily considers the risk and need principle of the offender, besides the motivation,
and treatment and responsibility principles .
The EBP focus for education area is improving the quality of research and evalua-
tion on education programs and practices, and hence, the information diffusion in the
educational research field to be used by professionals or policies creators .
We generalize the EBP steps as follows:
1. Transforming the need for information into a question that can be answered;
2. Identifying the best evidence to answer the question;
3. Critically analyzing the evidence to answer:
• Is it valid (appropriate methodology and proximity to the truth)?
• Is it relevant (size and significance of the observed effects)?
• Can it help (applicable in professional practice)?
4. Integrating critical analysis with professional skills and the values and cultural
aspects of the actor of the problem answering:
• How much the evidence can help the actor in particular (expectative from inter-
• Is it adaptable to actor’s goal and preferences (similarity between sample of the
study and profile of the actor)?
• How much safety can be expected (test result present into document)?
5. Evaluating the efficiency and effectiveness of the results of each step for future
In step 1, the question is usually formulated with components called PICO: Problem
(and/or actor), Intervention, Comparison of interventions, and Outcome . In step 2,
the identification of the best evidence is made, mainly, considering type of study,
information source provider, data sources considered on study, types of intervention
presented, results, references and sample. In steps 3 and 4, the questions were adapted
from Heneghan and Badenoch  and the answers of them represent contextual
information that supports decision-making.
2.3 Case-Based Reasoning
Case-Based Reasoning is a kind of reasoning that search the solution for a certain
problem through a comparative analysis of previous realities with a new similar real-
ity occurred .
One case is the primary knowledge element structured in combination of problem
features and actions associated with its resolution. It comprises three main parts: (i)
description of the problem - in general, presents the characteristics of their
344 E.C. Lopes et al.
occurrence, intentions or goals to be achieved, and constraints (conditions that must
be considered), (ii) solution - express the derived solution to the problem, and (iii)
result - corresponds to a feedback of what happened in consequence of the imple-
mented solution, including no success. The case can be enriched with other informa-
tion, such as solving strategies (or a set of steps), justifications for the decisions that
were taken, and system performance when handling the Case Base .
CBR systems use a structure for representation and organization of cases, called
Case Memory, which is formed by the Case Base and the mechanisms for access to
the base. To be more easily and quickly retrieved, the cases are indexed according to a
set of characteristics that represent an interpretation of a specific situation. Indexing is
an instrument whose function is to guide the similarity among cases .
Aamodt and Plaza  established the basic cycle of CBR processing that can be
described as: given a problem, obtain relevant previous solutions (retrieve), adapt
them for the current problem (reuse), propose a solution with validation (revise), and
store the new case (retain) with its solution.
Thus, CBR is not only a computational technique, but also a methodology for
guided decision making .
3 Cases, Evidences and Context to Support Decision Making
In order to integrate the three concepts and applies its in a guide of decision making,
firstly, we extend the conceptual framework presented in Lopes et at. , which
represents evidence and context, to incorporate case structure justified by evidences.
This integration facilitates the reuse of evidence already applied, and consequently,
the basic cycle of RBC needed to be adapted, as will show in this section.
3.1 Incorporating Case to a Conceptual Framework
The primary aim of this conceptual framework is to provide a class structure that
represents information related to EBP procedures, while taking into consideration
information about its decision-making context. The domain analysis was done in
juridical, medical and educational environments, and includes: bibliographical re-
search, specific legislation research, analysis of real cases collected and interviews
with decision-makers. The Figure 1 presents the conceptual framework extended.
The classes that involve context are based on definitions given by Vieira .The
focus is treated as an association of a task with an agent, which has a role in problem
ContextualEntity represents the entities of the application conceptual model and is
characterized by at least one contextual element. A contextual element is a property
that can be identified by a set of attributes and relationships associated with a Contex-
tualEntity . The association between Focus and ContextualElements determines
what is relevant for a focus.
Characteristics attributed to the dimension (ContextType) and the method of
acquiring contextual elements (AcquisitionType) are considered in the framework.
Contextual sources (SourceType) may be internal or external to the decision-making
environment (e.g., the patient’s medical records, a document with evidence obtained
Using Cases, Evidences and Context to Support Decision Making 345
Fig. 1. The conceptual framework extended to integrate cases, evidences and context
About EBP and its representation classes, the starting point is the observation of a
problem motivated by an actor to be decided by an agent.
Each problem is associated with an inquiry that is initiated by a formulated ques-
tion (see step 1 of the EBP procedures in Section 2.2), and completed with a self-
evaluation of the research performance and suggestions for the future (see step 5 of
the EBP procedures), whose information is instantiated in the Research class. Each
domain in which EBP is applied has a list of different types of questions (Question-
Type). For example: "diagnosis" and "prognosis" in the medical area, "drug testing"
and "occurring disorders" in the area of crime prevention, and "educational research"
During the research for evidences, several searches can be performed to retrieve
documents. For the Seek class, the expression and the type of search (SeekType) -
“title”, “author” or “subject” - must be present. InformationSource represents the
independent research groups that generate documents with evidence, such as Coch-
rane Collaboration (medical area) and Campbell Collaboration (areas of education
and crime prevention). Springer Verlag is not generating evidence, but has held
documents with evidence.
346 E.C. Lopes et al.
Each document presents a type of study that can be in all domains (e.g. systematic
review, case study) or more present in the specific domain (cohort - in the medical
area; narrative - in crime prevention; action-research - in education). Systematic re-
view and meta-analysis are studies of second degree; the remains are of first degree
. The association between type of study and degree of study is represented by type
attribute (Study_Degree) in the Document class.
In the medical area, Evidence-Based Medicine Guidelines are clinical guidelines
for primary care combined with the best available evidence. The framework is ex-
tendible from the perspective of using guidelines adapted as a type of study.
After selecting the found evidences, the agent (decision maker) will choose the one
that seems the most appropriate (step 2 of EBP), which is instantiated in the Evidence
The result of the critical analysis – or in other words the validity, relevance and
applicability of the best evidence (step 3 of EBP) – corresponds to contextual informa-
tion. Relevance is a contextual element in Document, while applicability (practical
utility) is in Evidence. Thus, Document and Evidence are specializations of Contextu-
The Intervention class is the result of an association among the Problem, Actor and
Evidence classes. It contains a description of a decision made (intervening solution)
where information about associated classes have been considered including prefer-
ences, values and cultural aspects (conduct, behaviour for example) of the actor with
the problem presented (step 4 of EBP). A preference is a contextual element and
hence Actor is a specialization of ContextualEntity.
About case structure, in highlighted in Figure 1, The Case class is composed of
Problem and Intervention component classes and aggregates Result and Learning
classes. Intervention represents the case solution and it has a relation with Evidence
class that corresponds to justification to the solution. Result and Learning classes are
extensions of the framework. Result contains information about obtained outcome and
the analysis of the outcome. As a research is part of the learning process, the Research
class is a component of the Learning class and detains the descriptions for the solu-
tion procedures and for the search performance in the Case Base.
3.2 Adapting EBP Procedures and Context to CBR Cycle
Aiming to incorporate EBP procedures into the CBR cycle, we have considered some
points. In order to consider the reuse of evidences in domains that use EBP, the clas-
sical case structure needs to be extended to incorporate the justification of the solution
(the research evidences found), becoming a case justified by evidence. Besides, in-
formation about learning in EBP (retrieved document history, decision maker’s self-
evaluation and suggestions to the future) must be presented in the new case structure.
In relation to contextual elements, it will be considered information related to ac-
tor’s profile (e.g. behavior) and preferences (e.g. treatment options and durability in
Evidence-Based Medicine) and decision maker’s profile (e.g. expertise).
In consequence, the CBR cycle needs to be adapted according to the Figure 2,
which contains the following activities:
Retrieve and Filter - Obtain cases with problems similar to a new case and apply
filtering mechanisms to show solved cases by decision makers with similar profiles
(e.g. expertise or areas of interest) to the new case decision maker’s;
Using Cases, Evidences and Context to Support Decision Making 347
Fig. 2. The adapted CBR cycle
Transform - It aims to build the PICO question to the presented problem. It corre-
sponds to the step 1 of the EBP;
Site Retrieval - it contains: (i) the document retrieval with found evidences regarding
the keywords of the PICO question applying Information Retrieval techniques with
similarity metrics and ranking schemes; and (ii) the selection of the best evidence
considering information from the source provider and aspects presented in the docu-
ment (e.g. type of study, evidence, data sources, etc.). It corresponds to the step 2 of
Evaluate Critically – during this activity, the methodological aspects and relevance of
results for studies presented into document must be evaluated. Besides practical ap-
plicability of the best evidence, must be considerate for corresponding to step 3 of
Integrate – The indication of evidence adaptation, based on intervention proposed in
documents, and the measures of expectative and safety of this adapted intervention,
must be integrated to actor’s goal and preferences – it is the step 4 of the EBP.
Reuse - it is related to the construction of new solutions. To this it is necessary to
consider: (i) similar cases solutions (or part of them); (ii) history of actor’s same cases
(recurrence turns the solution more complex); (iii) the evidence integrated with ac-
tor’s goal and preferences; and (iv) the actor’s profile.
Revise - Evaluate and test the solution recommended to determine their correctness,
utility and robustness.
348 E.C. Lopes et al.
Self-evaluate - It is a self-evaluation about the PICO question building, the evidences
search performance and the choice of the best evidence – it is the last step of the EBP.
Retain - Corresponds to the case learned that is added to the Case-base. Suggestions
for future researches, comments about strategies (or procedures) for the solution and
system performance in Case-base (Retrieve and Filter activity) can be registered.
4 Applying the Conceptual Framework Extended in the Area of
In this section we present the application of the conceptual framework extended. At
first, we represent cases, evidences and context integrated and, in sequence, we show
an implementation of case justified by evidence for the area of Crime Prevention.
The Pernambuco state court (Brazil) was chosen for application of the framework
extended because of its pioneering work on “restorative justice” and “therapeutic
justice”, themes inherent in Evidence-Based Crime Prevention. The main require-
ments are: to judge cases through judicial sentences, and to make interventions based
on support programs to the involved participants with the objective to avoid recidi-
vism. This work corresponds to the second requirement.
4.1 Integrating Case, Context and Evidence in the Area of Crime Prevention
The instantiated framework extended is presented in Figure 3 enriched with the
stereotypes <<ce>> and <<contextualEntity>>, corresponding respectively to the
concepts ContextualElement and ContextualEntity.
In respect to the focus, each EBP procedure corresponds to a task. The following
tasks were identified: (i) "make a juridical question”; (ii) "find the best juridical re-
search evidence" based on the designation of juridical site providers, types of study
and search expressions associated with the given question; (iii) "make a critical analy-
sis of the best evidence found"; (iv) "integrate the best evidence found with the values
and preferences of the participant with presented problem"; and (v) “do a self evalua-
tion of the judge’s performance” to measure all the tasks of EBP. "Translator" and
"designer" are the respective roles for tasks (i) and (ii); "intervenor” for the task (iv),
while "evaluator" for the other tasks. To model the first focus, we consider the Juridi-
calFact and Participant classes and the Victim and Defendant subclasses, which ap-
pear highlighted in the illustration. Figure 3 represents the modeling of all investi-
The association between Judge and JuridicalFact brings up the judges that make
interventions of supported programs. The characterization of the problem is given
through the constitution of the juridical facts and the circumstances that motivated the
offender being represented in the JuridicalFact class. To facilitate the information
retrieval based on problems, key terms related to the juridical fact will be instantiated
in the JuridicalFact class. The offender’s personal data are represented in the Defen-
dant subclass inherited from Participant. In several cases, the presence of victims
occurs. Thus, Defendant and Victim are specializations of Participant.
Using Cases, Evidences and Context to Support Decision Making 349
description : string
summary : String
suggestedInt : String
question : String
type : QuestionType
historic : sting
suggestion : String
description : String
<<ce>>safety : double
name : String
login : String
password : String
name : String
conduct : String
behaviour : String
needs : String
<<ce>>abilities : double
<<ce>>risk : double
expression : String
location : String
ttle : String
author : String
keyword[1..9] : String
type : Study_Degree
name : String
homePage : String
Fig. 3. The framework extended applied to the area of crime prevention
The formulation of a question is based on data from the participant, possible inter-
ventions (programs like parent counseling, to support victims of crime, cyber abuse in
children and adolescents, etc.) and desired results. The question and its corresponding
type are instantiated in the JuridicalResearch class. The historic attribute in this class
should include number of documents accepted and rejected. Searches for evidences
should mention the period of validity for the documents requested in each reliable site
(start and end).
For the ResearchedDocument class, the required attributes (besides the contextual
elements) are: location (URI / URL), title, author, keywords, publication and sample
of the study (participants, age interval, geographic and temporal aspects, etc.).
Searches for secondary studies should be conducted on Campbell Collaboration’s and
350 E.C. Lopes et al.
Springer websites. Primary studies should be obtained on the websites of Courts (fed-
eral or state) and in rely electronic journals in the country (JusNavigandi, National
Association of Therapeutic Justice, etc.). The homePage attribute value is the refer-
ence to the JuridicalEvidProvider class that holds judicial evidences.
Regarding the Evidence class, it should contain a summary of found evidences and
the suggested interventions contained in the document. Information about priority
solution that contains the proposals of evidence-based intervention must be presented
in the RestorativeIntervention class.
Regarding contextual entities and elements for the Crime Prevention area were
identified, as follows:
1. validity (ResearchedDocument) - indicates whether the document should be
selected based on its quality and the methodological rigor;
2. relevance (ResearchedDocument) - indicates whether the set of outcomes in the
document, often presented in statistical format, is consistent and significant;
3. applicability (Evidence) - indicates whether the evidence has practical utility in
4. sample (ResearchedDocument) – denotes contextual aspects about the study pre-
sented and serves to match with participants’ contextual information;
5. abilities (Participant) – represents the actor’s skills (profile), and is used to find
mutual affinities with intervention programs (e.g. revenue);
6. availability (Participant) – registers the availability preferences, measured in days
and shifts. A participant with a good availability chart has more alternatives and higher
chances of fulfilling the intervention on the schedule defined by the Judge;
7. adaptability (RestorativeIntervention) – indicates the degree of coherence in the
application of the evidence for the profile (including abilities) and preferences (avail-
ability) of the participant;
8. safety (RestorativeIntervention) - denotes the percentage of safety that has the deci-
sion maker to apply the specific evidence into a particular participant;
9. expectation (RestorativeIntervention) - refers to the percentage of expected support
from the use of evidence in relation to the participant;
10. expertAffinity (Judge) – identifies a relation of expertise from the Judge profile on a
given subject matter (e.g. homicide) representing mutual affinities among judges;
11. subjectSimilarity (ResearchedDocument, Judge) – it refers to the percentage of
similarity between keywords in a document and subjects of interest for the Judge;
12. recurrence (Defendant, JuridicalFact) – indicates if the defendant is a primary
defendant or not. It increases the complexity of the problem;
13. risk (Defendant) – it comes from juridical and psychosocial evaluations (profile).
Behavior data, conduct and fact description, especially for recurrent cases, are bases
for measuring the degree of risk;
14. complexity (RestorativeIntervention) – it comes from the juridical evaluation and
represents the degree of difficulty that the judge had in solving the case and indicates
the intervention program. Recurrence and risk increase this element;
15. situation (JuridicalFact, RestorativeIntervention) – indicates whether the problem
is ongoing or solved.
Using Cases, Evidences and Context to Support Decision Making 351
description : string
summary : String
suggestedInt : String
question : String
type : QuestionType
historic : sting
suggestion : String
description : String
<<ce>>safety : double
Fig. 4. Case representation model for the area of crime prevention
Considering cases, it was necessary to represent the case justified by evidence ac-
cording to the Figure 4, because the case-based reasoner depends heavily on the case
structure and its contents to operate.
The JuridicalFact and RestorativeIntervention classes represent case problem and
solution, respectively. The intervention (description in RestorativeIntervention class)
is based on evidence that serves of justification (summary) to the solution. Outcome
description and discussion are preview in Result class. Procedures of the interventions
(intervProc) and performance in search in Case Base (casebasePerf) are presented in
Learning class that is complemented with data of the JuridicalResearch class.
We present the aspects related to the implementation of an example adapted from a
real case in the crime prevention domain involving an alternative penalty - a model
for infractions that are of minor and moderately offensive potential (e.g. contravention
or illegal weapon possession). It deals with a new modality, face-to-face restorative
justice, in which a victim that suffered an assault with a weapon from an alcoholic
offender receives a support.
In order to illustrate the technical viability of the proposed integration, a prototype
was developed, using the Java language. It interacts with the framework JColibri 2.1
, which includes filter-based retrieval (FilterBasedRetrievalMethod) for cases that
satisfy a query expressed in a subset of SQL. JColibri is an open source implementa-
tion that provides expansions for thesaurus inclusion, textual and semantic researches,
techniques for information extraction, among other advantages. For the cases storage
we used the database manager PostgreSQL version 8.3.
Figure 5 presents some arguments to find cases justified by evidence in the Court’s
Case-base. The similarity functions used in JColibri were, respectively, Maxstring
(juridical fact description) and Equal (juridical fact participant – we used the initials
of the authors’ name). We applied Salton’s cosine formula  used in Information
Retrieval for keyword similarity between the formulated query and the retrieved
documents with evidence.
We made two information retrievals. Without use contextual elements, the results
with several cases are presented in Figure 6a (the expertise d.c means “drug crimes”,
c.a.c - “crimes against child” and hom. - “homicide”). When it was used contextual
information parameters as filter, a considerable reduction was observed in the selected
cases (see Figure 6b). The Judge’s expertise in the new case is “drug crimes”.
352 E.C. Lopes et al.
Fig. 5. Data for case retrieval with filtering mechanisms
Fig. 6. Retrieved cases justified by evidence: without filter (6a) and with filtering (6b)
Fig. 7. Data for searching evidences in Springer Verlag’s database
Using Cases, Evidences and Context to Support Decision Making 353
Fig. 8. Evaluate the best evidence found
Therefore, analyzing the presented cases, the judge did not considerate that they
were sufficient to support solution and he searched for evidences on the Internet.
The research began with the question containing: the problem and victim (woman
with a psychological problem who was assaulted), intervention (face-to-face ses-
sions), comparison of interventions (face-to-face sessions and conventional processes)
and outcome (beneficial effects). The sources Campbell Collaboration and Springer
Verlag were chosen and their home-pages were obtained. Figure 7 show data for the
second search regarding documents published between 2005 and 2010.
As shown in Figure 8, the document with the best evidence found was evaluated
and its information was manually extracted from web sites and filled in the form.
The integration of the best evidence found with goal and preferences is presented
in Figure 9.
Fig. 9. Evaluate the integration between evidence with goal and preferences of the victim
354 E.C. Lopes et al.
Fig. 10. Case generation
Data from the victim were informed and are compatible with the best evidence
found. To conclude, the case generation is presented in Figure 10. To do this, the
decision maker needs to inform the intervention, results and data related to the learn-
ing process (whether relative to the EBP or to the solution procedures and Case Base
performance). The victim agrees to participate in face-to-face meetings with the of-
fender, since that in previously scheduled time and with the presence of authorities.
This case is justified because many of the presented defendants suffered from
violence in the past and the crime victims could turn into offenders in the future .
5 Related Works
In this section we present some related work on the combination of themes involving
EBP, Context and CBR.
Dobrow et al.  emphasize the treatment of evidence with context. In a theoretical
approach about Evidence-Based Decision-Making for health policy, the authors pre-
sent a conceptual framework regarding the role of context in the evidence introduc-
tion, interpretation and application for decision-making support.
Kay et al.  describe ONCOR, an ontology- and evidence-based approach ap-
plied to contexts. They provide an approach to build ontology of places, devices and
sensors in ubiquitous computing in building environment. Locations, activities, ser-
vices and devices are considered in ONCOR in order to treat context history to model
indoor pervasive computing places.
Using Cases, Evidences and Context to Support Decision Making 355
A work combining PBE with RBC presents a knowledge-based system that inter-
acts with the Web, called CARE-PARTNER . Its purpose is to support users in
tasks involving clinical care of cancer patients who have undergone transplants. The
system applies reasoning about knowledge sources of expert committees, cases and
guidelines for clinical practice. CARE-PARTNER also considers negative feedback
for learning effect.
In , the authors developed a music recommendation system in CBR that uses
users’ demographics and behavioral patterns considering also his/her context. The
system identifies the user and collects the weather data from a web service. In se-
quence, the system retrieves the users (similar to the user identified) who listened to
music in the same context (profiles and weather) to select music for recommendation.
This work uses Database and Case Bases (available music, user’s profile and listening
history, respectively), and data from the Web.
These related works consider a combination of themes and were developed for
specific domains. But, none of them has the perspective of integrating the three con-
cepts and providing extension for several domains.
6 Conclusions and Future Work
This article proposes the extension of a conceptual framework that facilitate the de-
velopment of applications centered in EBP, with the consideration of context, to in-
corporate the case concept and the reuse of solutions justified by evidences previously
applied for domains that use EBP. It also proposes the incorporation of context and
EBP procedures to the CBR processing cycle aiming to support decision making. To
do this, it was necessary to represent the integration of context and evidence into a
classic case structure.
The classes of the framework extended and the integration representation were pre-
sented and experimented in the area of crime prevention. Contextual information
related to the EBP in this domain was modeled.
With a practical implementation for the Pernambuco state court, Brazil, we showed
how CBR techniques and EBP procedures can be used to support Judge’s decision
making. Besides, we observed that using contextual information in Case Base makes
retrieval and filtering mechanisms more effective.
Future researches encompass: (i) the incorporation of mechanisms to support group
decision making; (ii) the creation of a semi-automatic Evidence-Oriented Information
Extractor; and (iii) the development of a computational tool for risk assessment.
1. Loriggio, A.: De onde vem os problemas. Negocio Editora, São Paulo (2002)
2. Friedland, D.J., Go, A.S., Davoren, J.B., Shlipak, M.G., Bent, S.W., Subak, L.L., Mendel-
son, T.: Evidence-Based Medicine: A Framework for Clinical Practice. McGraw-Hill,
3. Sackett, D.L., Straus, S.E., Richardson, W.S., Rosenberg, W., Haynes, R.B.: Evidence-
based medicine: how to practice and teach EBM. Elsevier Health Sciences, New York
356 E.C. Lopes et al. Download full-text
4. Pal, S., Shiu, S.C.K.: Foundations of Soft Case-Based Reasoning. Wiley-Intersciense
Publication, New Jersey (2004)
5. Dobrow, M.J., Goel, V., Upshur, R.E.G.: Evidence-based health policy: context and
utilization. Social Science & Medicine 58(1), 207–217 (2004)
6. Vieira, V.: CEManTIKA: A Domain Independent Framework for Designing Context-
Sensitive Systems. Doctorate thesis, Federal University of Pernambuco, Brazil (2008)
7. Bunningen, A.: Context Aware Querying - Challenges for data management in ambient
intelligence. Doctorate thesis, University of Twente (2004)
8. Dey, A.K., Abowd, G.D.: A Conceptual Framework and a Toolkit for Supporting the
Rapid Prototyping of Context-Aware Applications. Human-Computer Interaction (HCI)
Journal 16(2-4), 97–106 (2001)
9. Henricksen, K., Indulska, J.: Developing Context-Aware Pervasive Computing Applica-
tions: Models and Approach. Pervasive and Mobile Computing Journal 2(1), 37–64 (2006)
10. Brézillon, P.: Context modeling: Task model and practice model. In: Kokinov, B.,
Richardson, D.C., Roth-Berghofer, T.R., Vieu, L. (eds.) CONTEXT 2007. LNCS (LNAI),
vol. 4635, pp. 122–135. Springer, Heidelberg (2007)
11. Thomas, G., Pring, R.: Evidence-Based Practice in Education. Open University Press, UK
12. Warren, R.K.: Evidence-Based Practice to Reduce Recidivism: Implications for State Judi-
ciaries (2007), http://works.bepress.com/roger_warren/1
13. Heneghan, C., Badenoch, D.: Evidence-Based Medicine Toolkit. Blackwell Publishing
Ltd., Oxford (2006)
14. Kolodner, J.: Case-Based Reasoning. Morgan Kaufmann Publishers, San Mateo (1993)
15. Aamodt, A., Plaza, E.: Case-Based Reasoning: foundational issues, methodological varia-
tions and system approaches. AI Communications 7(1), 39–59 (1994)
16. Belecheanu, R., Haque, B., Pawar, K.S., Barson, R.: Decision Support Methodology for
Early Decision Making in New Product Development: A Case Based Reasoning Approach.
University of Nottingham (1999),
17. Lopes, E., Schiel, U., Vieira, V., Salgado, A.: A Conceptual Framework for the Develop-
ment of Applications Centred on Context and Evidence-Based Practice. In: Filipe, J., Cor-
deiro, J. (eds.) ICEIS 2010. LNBIP, vol. 73, pp. 331–347. Springer, Heidelberg (2011)
18. Bello-Tomás, J.J., González-Calero, P.A., Díaz-Agudo, B.: JColibri: An object-oriented
framework for building CBR systems. In: Funk, P., González Calero, P.A. (eds.) ECCBR
2004. LNCS (LNAI), vol. 3155, pp. 32–46. Springer, Heidelberg (2004)
19. Salton, G.: Automatic Information Organization and Retrieval. McGraw-Hill, New York
20. Sherman, L.W., Strang, H., Angel, C., Woods, D.J., Barnes, G.C., Inkpen, N., Bennett,
S.B., Rossner, M.: Effects of face-to-face restorative justice on victims of crime in four
randomized, controlled trials. Journal of Experimental Criminology 2(3), 407–435 (2006)
21. Kay, J., Niu, W.T., Carmichael, D.J.: ONCOR: Ontology- and Evidence-based Context
Reasoner. In: Intelligent User Interface – IUI 2007, pp. 290–293. ACM, Honolulu (2007)
22. Bichindaritz, I., Kansu, E., Sullivan, K.M.: Case-Based Reasoning in CARE-PARTNER:
Gathering Evidence for Evidence-Based Medical Practice. In: Smyth, B., Cunningham, P.
(eds.) EWCBR 1998. LNCS (LNAI), vol. 1488, pp. 334–345. Springer, Heidelberg (1998)
23. Lee, J.S., Lee, J.C.: Context Awareness by Case-Based Reasoning in a Music Recommen-
dation System. In: Ichikawa, H., Cho, W.-D., Chen, Y., Youn, H.Y. (eds.) UCS 2007.
LNCS, vol. 4836, pp. 45–58. Springer, Heidelberg (2007)