1/3/2008
1. Methods and Techniques for developing Research projects A possible vision about Theories on the development and characterization of Scientific Research
Eugénio Oliveira- 2007/2008
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
1. Abstract of my Lectures Understanding about Science Scientific Methods and Techniques Writing scientific papers Evaluating scientific papers Consulting scientific Data Bases and searching for information ( Dr. Ana Azevedo)
2. Abstract of your Work Criticizing a scientific paper Organizing a workshop Writing a scientific paper Presenting a scientific paper Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
1
1/3/2008
1.1 Basic Motivations: Issues on Scientific Methods
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Source: GETA and Graduate School of Electrical and Communications Engineering course: Introduction to Research Methodology, Aarne Mammela, VTTHelsinki University of Technology,
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
2
1/3/2008
• Basic Characteristics: • Cumulative versus reformulating always from the principles • However: •Thomas Kuhn (Harvard, Berkley, Princeton, MIT until 91. +96) “The Structure of Scientific Revolutions” 1962. • He points out the role in scientific research of the emergence of NEW PARADIGMS. •A PARADIGM is defined as: Scientific discoveries, which universally recognized during a certain period of time, create a new model for both problems and their respective solutions to be used by a community of practicians ans experimentalists Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Paradigms support theories based on concepts, phenomena and techniques for helping on explaining new facts or information .
• Science progresses through successive ruptures depending on the current accepted paradigms. Paradigms correspond to “different views” over the reality which are influenced by aspects other than scientific ones. Wittgenstein and duck-??????…
L.Wittgenstein “Tractatus Logico-Philosophicus
Wttgeinstein’s Poker”, David Edmonds, J. Eidinow Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
3
1/3/2008
• Scientific interpretation (and perception) is mostly affected by Social as well as Psychological factors. • Does it lead to Irrationality dominating Science? No ! • A good discussion on prevalence of science can be found in: • Kuhn, T.S. “Black-Body Theory and the Quantum Discontinuity”, 1987. However: Paradigm change non-cumulative Science (sometimes) • Example of important Paradigm changes: • Solar System Ptolemy's vision changed by Copernicus vision • Going from Newtonian Phisics to Quantic Physics (Uncertainty Versus Determinism..) and Relativity (SpaceTime relationship). Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Karl Popper (London S. E., + 94) • “The Logic of Scientific Discover”, Basic Books, 1959 (1ªEd Ing) • Conjecture and Refutations, Harper & Row, 1968 • Idealistic. He developed a Scientific Method as an Epistomology (a way of “knowing”) independently of the intrinsic value. • T. Khun e Feyerabend (1993) criticize this point of view showing how science is guided by value and, therefore being non-idealistic. • This is a debate on “Consciousness over Existence”
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
4
1/3/2008
• “Critical Rationalism” proposal defends “falsifiability” as a criterion for demarcating (separating) science and non-science Induction Baaaad!!! - Falsifiability gooood!!! • Popper argues that scientific theory will always be conjectural and provisional. It is not possible to confirm veracity (the truth) of a theory by means of simply recognizing that the results of a preview already done based on that very same theory, hold. That Theory should only be seen as a theory not (or not yet) contradicted by facts.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Verificationism: a sentence should be either verifiable or falsifiable Critics addressed to Verificationism:
Generalization: specific instances lead to universal assumptions Observation X, being an instance of Theory T Increases the probability of Theory T being correct
The “Black Crow paradox”: • The “Black Crow paradox”: by Hempel: “All Crows are Black" is logically equivalent to “Everything which is Non-Black is not a Crow". ∀x(Cx Bx) ∀ x(¬Bx ¬Cx) • Thus, If ∃ x(Cx ∧ Bx) reassures that ∀ x(Cx Bx), Then ∃ x(¬Bx ∧ ¬Cx) reassures that ∀ x(¬Bx ¬Cx) which, being equivalent to ∀ x(Cx Bx), this last implication would also see its probability of being True increased. Eugénio Oliveira- 2007
MSR-Metodologias MICMethodologies defor Investigação Scientific Research Científica
5
1/3/2008
• Crítics addressed to Verificationism:
• Generalization: specific instances lead to universal assumptions
• “There are Apples (non-Crows) not Black” Increases the probability that the sentence “All Crows are Black” is True • If verificationism (generalization) was acceptable, any trivial evidence would support a Theory There are attempts to overcome this paradox.. “falsifiability” proposed by KP states that keeping just the Existential assertion permit us to know that if another existential proposition that contradicts it, arrises, this falsifies the Theory T !!
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
•
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
6
1/3/2008
A An exploration Enterprise: Christopher Columbus – the “Explorer” Motivation: ambiguous… Problem: a new way to “India” Hypothesis: over the Atlantic and not around south Africa Experiment: Yes he have done it !! Conclusions: ????? Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
A research: discover new knowledge • basic research (no specific application in mind)
• applied research (ideas into operational form) • development: systematic use of the existing knowledge
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
7
1/3/2008
A
Science, Technology and Engineering [Jain97] • science: organized or systematic body of knowledge • technology: application of scientific knowledge for practical ends in engineering, medicine, agriculture, etc. • •natural sciences and engineering sciences differ in the object of study: • natural sciences: objects in the nature • engineering sciences: objects (products, services, methods) not found in the nature, using results of mathematics and natural sciences Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
A Science, Technology and Engineering [Jain97] Verification, confirmation, and justification are synonymous terms in philosophy of science. The opposite is falsification or refutation.
R. K. Jain and H. C. Triandis, Management of Research and Development Organizations: Managing the Unmanageable. John Wiley & Sons, 1997.
Requirements for success • You have to be: • analytical and curious, • autonomous and collaborative • criticism and tolerant Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
8
1/3/2008
A How does a researcher work? •
• make always notes in a “notebook” • make summaries on what has been learned. • make plans for the future all the time (outlines, roadmaps) • discuss, ask questions and argue (criticism)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
A
Learning carefully by induction (bottom up, generalization from examples to models) [Felder88] • Defending theories by deduction (top down, from models to results) •
Break the problem down and then generalize the results (divide and conquer) R. M. Felder and L. K. Silverman, “Learning and teaching styles in engineering education, ”Engineering Education, pp. 674681, April 1988.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
9
1/3/2008
A
1.2 Supporting Theories
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Basic Characteristics • Research involves Methodologie(s) and Techniques • A- Methodology includes the study of several methods that are applicable to a class of problems • i.e., the set of processes to conduct each specific Research Project • B- Techniques refer to the means and specific Tools enabling relevant information acquisition, the respective data analysis, as well as the inferences that can be made.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
10
1/3/2008
• A- Methodologies • depending on the MOTIVATION: • ‘Pure Research’:
• It contributes to a deep theoretical understanding and for a more abstract formulation of the phenomena.
• ‘Instrumentalist Research’: • It contributes for making human intervention in Real world environments, more effective. • Two major sub-categories are identified: • ‘Applied Research’ • ‘Problem - Oriented Research’
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• ‘Apllied Research’.
• Starting from a technology (devices, specific techniques, both) use them for dealing with processes (physic, organizational, social, individual,...)
• ‘Problem- Oriented Research’.
• Starting with the problem description and then looking for the adequated techniques to solve it.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
11
1/3/2008
• Depending on the Theory underlying the Research: • ‘Descriptive’, of all significative aspects of the domain • ‘Explicative’, of the behaviour of all the phenomena • ‘Predictive’ of the future phenomena behaviour • ‘Prescriptive’, beyond the prediction it should also prescribe and apply Norms and processes in identified specific circumstances (Research in IS, Electronics and Telecommunications should be of this kind)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Depending on the tradition in that scientific area or Knowledge Domain: i) ‘Traditional Scientific Research’ which is more quantitative ii) ‘Interpretative Research’ which is more qualitative iii) ‘Engineering-oriented Research’
• iii) most appropriated for Electronics, Telecommunications and CS • But do not ignore the other approaches!
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
12
1/3/2008
1.3 Positivist (classical) Research method
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Traditional Scientific Research is based on positivist rational thinking:
• From observations we may build up Theories trying to explain what has been observed • Theories are expressed in Deductive form or using Axioms and Postulates which are then operated through Logics.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
13
1/3/2008
Do Not Forget :
• Scientific Theories, following K. Popper, are capable of generating Inferences which are, in principle, falsifiable through references to the real world. • Results coming from the process of testing hypothesis give feedback to the Theory, enabling verification of its trueness or detecting possible “ab-normalities” (T.Kuhn).
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Whenever the Observation comes from the Real World it is called empirical and it becomes relevant in two situations:
• during the formation of the theoretical knowledge through the so-called passive observations • during Hypothesis verification, gathering information in a more active way, although guided by that same theory.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
14
1/3/2008
• When we do not have a suitable Theory available, we have to pursue
‘exploratory research’ where experiments and interpretation are open, not guided by a Theory and making available Empirical Knowledge that may postulate other Theories. • IS and Telecommunications are disciplines (scientific areas) without a given stable theory. • They encompass applied subjects using a partial theory or co-opting a theory from “reference disciplines” such as: • Mathematics • Information Theory (C.Shannon) • Physics • “Theory of Organizations”, • “Management Sciences”, • “Computer Science”. Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
3 Theory of Disciplines of Reference
Theory Axioms Postulates Deductive Logic
Hypothesis
1
4
7
6
Research Project 5
Real World
Simulation
2
1- There exists a Real World that we can Perceive. 2- Knowledge is Organized in Theories making inferences possible 3- Applied Disciplines like IS or T rely on Theories of Reference of other Disciplines 4- A Research Project permits testing formal Hypothesis formulated through inferences expressed in the Theory. 5- In Empirical Research, Hypothesis derived from the Theory are directly cheked in the Real Worldl 6- Testing Hypothesis through Simulation 7- Results comming out of the Research give feedback to the Theory
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
15
1/3/2008
• Summarizing: In traditional (or conventional) Science, new Hypothesis are extracted from the existent Theory, they are tested and new results are added to that Theory.
• Thus, it implies the pre-existance of: • A theoretical Body of Knowledge • An explicit theoretical Framework to guide the research • Definition of what issues have to be investigated • What are the explicit Hypotheses that could be refuted (falsifiability) • Method for Applied Research and well defined techniques for testing the Hypothesis.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
1.4 Interpretative Research method is concerned with trying to understand lived experience and with how participants themselves make sense of their experiences. Therefore it is concerned with the meanings which those experiences hold for the participants. IRM is phenomenological in that it wishes to explore an individual’s personal perception or account of an event or state as opposed to attempting to produce an objective record of the event or state itself. http://www.psyc.bbk.ac.uk/ipa/ (Birkback- U. London) Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
16
1/3/2008
• Interpretative Method (or philosophy): • post-positivist • criticizes the (conventional) scientific “chimera” • declares the impossibility of objective observation • Observations and Interpretations depend on the Observer. • Addresses critics to both “hard Sciences” and Social Sciences, • Difficulties for Objectivity:
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
• Concepts, Factors, relationships that can’t be accessed • Researchers involvement in the research domain • Results depending on the researcher own perspective: - On selecting and defining the research (sub-) domain - On selecting a pre-existant theory - On setting the issues to be investigated - On setting the precise scope of the research - On selecting and defining the Variables to be instantiated - On measuring those variables possible values • Different attitudes on all those issues may lead to many different interpretations of the same observed phenomena.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
17
1/3/2008
Engineering-oriented Research Approach
1.5 Research in Engineering disciplines
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Engineering-oriented Research Approach • For this approach to research, technology is very important! • (devices, artefacts, practical techniques to be used) • At least up to a certain extent is the most suitable for research in Information Systems, Electronics, Telecommunications. • Research in IS and Telecomm is applied and problem-oriented. • Implying: - application of suitable technology, - conceptualization: requirements, models - prototyping, - implementation - demonstration of (new) technology (real world/sim) - Evaluation phase
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
18
1/3/2008
Engineering-oriented Research Approach
Observation and Theory [Wohlin99] • In engineering an hypothesis (defined in system specifications) is usually an idea of the relationship between the cause and effect (defined in system requirements) • Theoretical model is always only an approximation of observation in real world
C. Wohlin et al., Experimentation in Software Engineering: An Introduction, Springer, 1999.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Engineering-oriented Research Approach
In engineering we are fundamentally interested in how efficiently the basic resources are used for a needed performance.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
19
1/3/2008
Engineering-oriented Research Approach • From Dr. Leung Yee Hong Australia TC Research Institut
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques
1.6 Techniques associated to Research
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
20
1/3/2008
Research Techniques B- Techniques I.
Non empírical Techniques
II.
Scientific Research positivist Techniques
III. Scientific Research Interpretative Techniques IV. Scientific Research Techniques at the positivist/interpretative boundery V.
Scientific Research Engineering-based Techniques
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques I.
Non empirical Techniques depend on: • •
Artificially created data, or Conceptual thinking about abstractions
•
Includes: • conceptual research based on opinion and speculation. ( argumentative and dialectic analysis) • theorem proof: This applies to mathematical abstractions: • formal methods application, • induction, • mathematical abstractions • Models verification • Simulation: • Mapping a complex environment in a simplified Model.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
21
1/3/2008
Research Techniques
•
Other non empiric Techniques: • • •
•
Using hypothetic future scenarios Mapping a real scenario into Games ...
Reviewing the State-of-the-art and doing Meta-analysis
Eugénio Oliveira- 2007
Março 2007
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
FEUP-LIACC / Eugénio Oliveira
MSR- Methodologies for Scientific Research
22
1/3/2008
Research Techniques
Scientific Research positivist Techniques (applicable to IS and other Engineering disciplines):
II.
•
Techniques for forecasting applying • regression algorithms and • time-series based algorithms to extrapolate on historic data
•
Field Experiments : • Possibly isolating and controlling a set of phenomena to be looked at
•
Laboratory Experiments: • Creating artificial setups in which phenomena, factors and variables are isolated and controlled.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques
Scientific Research Interpretativist Techniques:
III. •
Research approach is Descriptive/Interpretative: Going from empirical observations to a limited rigour analysis. Need for controlling the researcher’s perspective: - Critical self- Examination - Revising all the postulates and biases - Varying Observations - Submission to peer reviewing •
Group Research: • frequent discussion with a group of people including who may be affected by the application (or the technology) • Use Collaborative Work Tools.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
23
1/3/2008
Research Techniques •
Action-based Research: - Researcher is like an “Agent” changing existent conditions to measure the respective reaction
•
Other Methods for Social Sciences(Ethnographic..)
In http://www.qual.auckland.ac.nz/ Michael Myers Ed. Ass.Inf.Sys. It is advocated qualitative research, for example, for IS. We also can do (but it is not recommended) R&D just through Case-study analysis Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques IV. Scientific Research Techniques at the positivist/interpretativist boundery Including: - Field Work : Data related to the object of study is directly observed by the Researcher in the original context - Questionary-based : Structured Data collection comming from interviewees -
Case Study: - Isolate and describe a relevant real-world situation, - Gather a collection of data from multiple sources
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
24
1/3/2008
Research Techniques
•
Secondary Research: •
Analysis starting from pre-existent documents (texts, papers, descriptions, memoirs).
•
Analysis is done under new perspective and using different techniques or procedures.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques v. Scientific Research Engineering-based Techniques (Informatics Engineering, IS, Electronic Engineering, Telecomm...) •
Divided into two categories: 1- Construction Technique: • Includes Design, Project, Implementation (or prototyping) of a system (ex: Computer program System, or physical device) - objective: -- explicitly testing an hypothesis or -- to solve a class of problems. - it implies: -- Existence of either a theory or a model for explaining the results (testing the hypothesis)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
25
1/3/2008
Research Techniques
Materials and Methods • in exact sciences materials include definitions, model, and related assumptions
• theoretical model is defined (preferably a mathematical model), including boundary conditions (define the environment) and initial conditions (define the initial values of the parameters of the model) [Losee01] J. Losee, A Historical Introduction to the Philosophy of Science, 4th ed. Oxford Univ. Press, 2001.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques • methods include: rules of analysis and rules of verification and validation • rules of analysis follow deductive reasoning, statistical analysis and approximations (analysis must often be replaced by simulations due to mathematical tractability problems) • the theoretical model is verified by comparing the results with reality (measurements with a prototype, known as hypothetical-deductive method)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
26
1/3/2008
Research Techniques • Results • results can be analytical (deductive), from simulation, or measurement results • it is important to present numerical results that verify (or not) your own thesis, • use analytical results in simple limited cases to obtain reference values • the value of the results depend on how well they can be generalized (bottom up approach will help you)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research Techniques Conclusions (constructive technique report) (1) “We have shown that [some] algorithms proved right…” (2) “Our approach is simpler to implement than previous algorithms and is practically feasible …. On the theoretical side, the greatest deficiency in our results is …” (3) “In a practical situation the advantages and limitations of our approach are…” (4) “These results can be extended in a number of directions…” 1. What is shown by this work and its significance. 2. Limitations and advantages. 3. Applications of the results. 4. Recommendations for further work.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
27
1/3/2008
Research Techniques
•
Destructive Techniques: •
To analyse a system, or a set of methods to gather new information about those technology procedures under analysis. - Recognized information and characteristics are then reused may be in a different way - Example: Reverse Engineering
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Phylosophy Underlying Research •
There are philosophies permitting the Classification of Research Methods through Paradigms or basic Epistemologies:
I. II. III. IV.
Positivist Interpretative (post-positivist) Critical (Constructivism)
At Harvard Business School they use just the three categories
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
28
1/3/2008
Philosophy Underlying Research •
Positivist Method: •
Assumes that reality is objectively accessible, it can be measured and described by the observer and measured by the same observer through their instruments and devices. Tests are made making predictions possible over future phenomena
•
IS research (also Telecomm research?) can be of the positivist kind (for IS, Orlikowski and Baroudi,1995) - Use math/logic formalisms - use quantifiable and measurable variables - Test an Hypothesis and make inferences from what is known.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Philosophy Underlying Research •
Interpretativist Method •
Assumes that accessing to reality is only possible through social constructions like language, consciousness and ontologies (sharing of concepts’ meaning).
•
Hermeneutics-based philosophy (based on the human understanding and text interpretation) as well as in phenomenology (describing, understanding and interpretation perceived phenomena) Proposes to abolish separation between "subject“ and "object", It opposes to XIX century positivist thinking.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
29
1/3/2008
Philosophy Underlying Research
•
Can IS and Telecomm studies influence or be influenced by the context?
•
Here, the focus is not as much on the variables, measurements and processes as it is on interpretation of what is being the results of the test, the measurement or implementation
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Philosophy Underlying Research •
Research Criticism: •
It is a radical enterprise once it tries to identify what constraints research: • socially, culturally, • Historically, politically, • economically, physically trying to emancipate science by eliminating causes for ignorance, alienation etc... Ex: Why is it not possible to do research on X? Ex; Why cannot we implement systems of type Y?
Habermas from Frankfurt School following the way of Adorno, Lukacs, Marcuse
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
30
1/3/2008
Philosophy Underlying Research
•
Junk Science (?) Stories of the Past Decade Dial “F” for Fear. Since the 1993 Larry King Live broadcast featuring a man suing a cell phone maker claiming his wife died from a cell phoneinduced brain cancer, many cell phone users have worried about phone safety. But studies failed to identify any risk.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Philosophy Underlying Research
•
Junk Science (?) Stories of the Past Decade
Powerline scare unplugged. Fears that electric and magnetic fields (EMFs) created by power lines and appliances caused cancer started in 1978. Parents worried about power lines over schools. Power companies worried about burying power lines. The National Academy of Sciences finally unplugged the scare in October 1996, concluding that no evidence showed EMFs presented a health hazard
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
31
1/3/2008
Philosophy Underlying Research
•
Junk Science Stories of the Past Decade The Lone-Tree Theory. It nearly took an act of Congress to get the researcher behind the notorious “hockey stick” graph, which purports to show a steep rise in global temperature in the 20th century following a millennium of stable temperatures, to release his publicly funded data and computer code. Among other dubious presumptions, the graph is derived from data that bases climate estimates for the entire 15th century on the tree ring measurements of a single tree.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Quality of Research
1.7 Quality of Research
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
32
1/3/2008
Quality of Research •
Quality depends on the Objectives to reach 1. To increase available (and also general) Knowledge • Objectives and Quality criteria are internal to the discipline • RIGOUR is very critical 2. Developing Computer Systems– (Instrumentalisme) • Objectives and Criteria are external (regarding R&D) • Relevance can be measured by the impact and is of the most importance • •
Relevance Versus Rigour Not antagonic but taken with different emphasis
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Quality of Research •
Computer Science (and other Engineering-like areas) tendency: • •
Until the 80ies conceptual (not empirical) papers dominated After the 80 ies Less theoretical papers, more instantiated, less general and Knowledge less cumulative
•
Case-studies based research is mainly descriptive concerning simple instances and then less scientific
•
Empirical research is now dominated by the descriptive approaches
•
Specific Models have limited capabilities for explaining a phenomen as well as little predictive power.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
33
1/3/2008
Quality of Research
•
Requirements for the Research in CS and TC •
Choose an appropriate Research Method:
•
•
Traditional/positivist, “interpretativiste” or Engineering-oriented Make explicit a Body of Theory sometimes coming from other adjacent disciplines (ex: Computer Science, Organization Theory, Information Theory, Physics, Statistics…)
•
Extend the Theory and Infer predictions to be checked (at least in the case of positivist/traditional research)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Scientific Methods, General
(following Björn Lisper, Datorteknik, Mälardalens högskola Västerås, Sweeden) http://www.idt.mdh.se/kurser/ct3340/ht03/ Theoretical methods: Create formal models (mathematics, logic) Define concepts within these Prove properties of the concepts Abstraction, hide details to make the whole more understandable (and to make it possible to prove properties of it) Proofs of properties by deductive methods Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
34
1/3/2008
Scientific Methods, General
Empirical methods: Perform experiments See how it turned out Draw conclusions Simulation: Start with a formal model at some "easy-to-understand" level Make "artificial experiments" in your computer Collect statistics and draw conclusions
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Scientific Methods
In Physics: Make hypotheses about the surrounding world (theory), observe it (experiment) Relate the result of experiment to the theory Adjust the theory if it doesn’t predict the reality well enough Theory is used to predict the future (e.g., if a bridge will hold for a certain load, or an asteroid fall down on our heads) Common pattern in Computing Science: The system is constructed to behave according to some theoretical model Deviations are seen as construction errors rather than deficiencies in the theory (hardware error, bug in OS, : : :) In both cases: the theory helps us understand and predict, but in different ways! Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
35
1/3/2008
Scientific Methods Theoretical vs. Empirical Methods in Computing Science Computing Science really has a “spectrum”, from “extreme constructivism” to a use of theory close to the one in physics: • “Extreme constructivism”: (ideal) programming language design: Formal semantics for the language, pure construction of model defining the mathematical meaning of each program Abstraction of details to make the meaning of the language simpler (for instance, assume that data structures can grow arbitrarily big) Implement the language according to the semantics
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Scientific Methods Theoretical vs. Empirical Methods in Computing Science
One can prove formally within the model that a program is correct –valuable! But the model does not cover all kinds of failures. E.g., hardware errors, or stack overflow (or an asteroid falling down on the computer)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
36
1/3/2008
Scientific Methods Theoretical vs. Empirical Methods in Computing Science
Extreme “physics” approach: performance modeling of complex computerand communication systems Extremely hard to make analytical calculations Simplified performance models, tested against experiments (e.g., long suites of benchmarks) Discrepancy leads to a modified theory, as in physics Often simulation (desire to evaluate systems before building them)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Scientific Methods Theoretical vs. Empirical Methods in Computing Science
In-between: algorithm analysis Build on some form of formal model for how the algorithm executs (meta-language with formal semantics), and some performance model (how long does a step in the algorithm take, how much memory is needed to store an entity)
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
37
1/3/2008
Scientific Methods Theoretical vs. Empirical Methods in Computing Science
In-between: algorithm analysis Given that the performance model is correct, one proves mathematically that the algorithm needs certain resources (time, memory) to be carried out But the performance model is often very approximate. Sometimes is possible to refine the performance model, but this can make it impossible to calculate the resource needs of the algorithm
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Quality of Research
•
Characteristics of the Research in CS and TC(?) • • •
Report the S-o-A. Prove you made an advancement in Knowledge Combination of Techniques trying to compensate one’s weaknesses with other one strengths •
Report about the object of study and the nature of the domain • Validate Data • Analysis using statistics • Extract relationships between variables • • • •
Praticable according to the resources Relevant results (at different levels: community,society...) Publish Ambitious
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
38
1/3/2008
Quality of Research
•
Characterístics of the research in CS: •
Research may have as an objective: • Building a Theory • Testing a Theory • Extending a Theory
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Quality of Research •
Other characteristics: •
Concerning Data: • Extensive • Representative • What is Data validity (ex. temporal, precision…) • Granularity: • Simple or agregate • Specific or large generic experiments
•
Temporal Horizon: • •
•
“Snapshot” of the domain at a specific point in time “Longitudinal cut section” (múltiple points in time)
Methods for colecting and analising data; quantification...
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
39
1/3/2008
Quality of Research
1.8 Challenges
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research in CS and TC Challenges •
Challenges: • There exists shortage of theory. Mainly theories in related disciplines • Phenomena are unstable: • Data may get old quickly, usually a snapshot that may vary Pay attention to the validity period • Organizations and Markets are always moving
•
Significant cultural variations: • Different national/regional cultures imply different speed/rhythm
•
When you are doing research (requirements, interaction, questionnaires, demos) you influence the domain
•
Pressure for publishing during the research action
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
40
1/3/2008
Research in CS and TC Challenges •
Pressure for technical instantiation:
•
Prototyping without a theoretical ground or well defined objective Economics and business models either not known or consolidate.
•
•
Impact of the automation and rationalization: • Depends on time dynamics concerns • Depends on who are directly affected
• •
Adopting a new technology may take long time Impact on other partners, markets, standards, tools…
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
Research in CS and TC Challenges •
Conclusions on Research actions in CS and TC • • • • • •
Consider tradoff Rigour Versus Relevance Identify challenges BEFORE starting doing research Be clear about conventions related with research to be done Select a significative sub-domain Define questions to be answered through that research Select the apropriate Research Method and Techniques (possibly a good combination of some) and justify the choice.
Whenever the selected aproach is instrumentalist, you need to be sure about the external quality, data validation and rigour as well as: Relevance
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
41
1/3/2008
Other Bibliography
• D. Sternberg, How to Complete and Survive a Doctoral Thesis. St. Martin’s Press, 1981.
Eugénio Oliveira- 2007
MSR- Methodologies for Scientific Research
42