Principles of Information Systems Session 04 Discovery and Representation
Feb 25, 2016
Principles of Information SystemsSession 04Discovery and Representation
Discovery and Representation
Chapter 3
2
3
OverviewLearning objectives
1. Introduction
2. Discovery
3. Knowledge elicitation and discovery
4. Other aspects of discovery
5. Representation
6. Discovery, representation and knowledge
7. Summary
Learning objectives• Explain why discovery and representation are related
• Describe several techniques for discovering information in unknown situations
• Explain why psychological techniques are often used in knowledge elicitation
• Use the repertory grid technique for eliciting an person’s constructed understanding of a an area of knowledge
4
Learning objectives
• Explain why the representation of something is different from the thing itself
• Describe several different categories of representation commonly used in informatics
• Discuss the importance of ensuring states of knowledge are matched when knowledge is being elicited and represented
5
Introduction
• The stuff of human knowledge, its symbolisation, its understanding within a larger framework of ideas, its systematisation, preservation and use all depend on representing what has been discovered, and recording it, at least for the time being.
• In this way findings or inspirations can be disembodied from their immediate context of discovery and shared with others across time and space .
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation and
knowledge7. Summary
6
Discovery and representationDiscovery is the process by which an idea is acquired…
…Representation is putting an idea into a form that can be conveyed to others
7
8
Why do we need representation in informatics?• You want to find out the requirements for an information system • You want to elicit the rare knowledge of an expert or the last living
speaker• You want to learn about hidden patterns of customer buying
behaviours• You want to discover genetic markers for a particular disease• …
… you are discovering the unknown or unexpressed and making it explicit through representation
9
Why do we need representation in informatics?• Communicate with a peer group• Communicate with others across time and space• Archive your learning, creations and observations for
posterity• Make a complex problem situation clearer• Model some aspect of the world
… you are communicating your discovery to others through representation
10
Conventions for representation• The conventions for representing data and ideas
have to be known by the person or source expressing the idea, and also recognised by those using the framework for receiving and using the idea, otherwise the message will be misunderstood.
• Informatics uses many representational technologies, with their own specific notations, signifiers and rules.
Conventions: agreed standard rules of usage by which something is consistently understood
11
Finding out something is intimately linked with how that understanding is
represented, and equally, representing something is
intimately linked with how it conveys new information.
Recap
12
Discovery
• Finding something that was previously unknown, or finding out something that was hidden, unspoken or obscure.
-Finding gold that was always there under the ground-Finding out what your new partner’s favourite song is-(Discovery also has technical meanings in fields such as law, but we won’t discuss them further here)
• Discovery implies a new idea has become available to an individual or community who can then store, process, learn from or otherwise use the idea.
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation and
knowledge7. Summary
13
How things are discovered
• Through personal experience
• Through reasoning from previously known information
• Through asking questions
• Through systematic research
14
Experience
• Knowledge or skills acquired through direct participation
• Our experience contributes to the framework of understanding we bring to learning new things and assimilating new discoveries
15
Reasoning
• Deduction – reaching a conclusion based on applying prior knowledge to observations
• Induction – working out general principles from instances or observations
• Abduction – hypothesising a general cause to explain a particular situation
A form of learning by moving between observation and
theories or explanations of those observations
16
Questioning
• Questions may be asked of experts, such as doctors, to find out specialised information
• Queries are formalised questions that may be addressed to databases, search engines or other computer based systems
Erotetics: the classical art of asking appropriate questions to
get to the heart of a matter
17
Open and closed questions
how was your day? was your
day:(a) good(b) bad(c) indifferent?
18
Questioning
• A key skill in informatics is the ability to ask good questions that elicit useful answers
-What information needs to be on the survey form?
-What should be recorded in the database?
-What do the users actually want from the new system?
• Research is finding out new information by systematic inquiry and investigation
19
Cyber forensics
• Investigation of crimes or misconduct, computer based or otherwise, involves information discovery and matching:
-Did the company director know the incriminating information? -Did that person pay for the music they are listening to? -Had the minister read that briefing? -Are the figures the same as those used in the financial audit?
• As informatics technologies become even more sophisticated, cyber forensics will continue as a growth area.
Cyber forensics locates, identifies and gathers digital
evidence, often in connection with crime investigations.
20
Information retrieval
• A branch of informatics concerned with discovering specific data in documents
-The ‘search’ or ‘find' buttons in web browsers, library catalogues or word processors
-Search engines
-Data mining techniques
• There is also a human skill in formulating appropriate queries - and assessing the results
21
Search engines
• Using search engines effectively is becoming an important information discovery technique
• Results are listed in order of relevance to your query, based on the search engine’s algorithm for finding and ranking web pages
-Boolean operators such as AND, OR and NOT allow you to filter your search effectively
- this idea is very widely used in informatics more generally, such as in database querying
22
Google circa 1960 (parody)
23
If knowledge already exists in represented form, techniques of
searching, discovery and reasoning can be applied
Recap
24
Knowledge elicitation and discovery• Not all knowledge is written down or articulated
• A lot of information or knowledge is “inside people’s heads” – it is implicit in:
-How they behave
-How they express themselves
-How they make connections between facts or observations
-How they feel about, understand or weigh up new situations
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation and
knowledge7. Summary
25
Knowledge elicitation
• Used in medical interviewing, prisoner interrogation, witness questioning by barristers, investigative journalism, chat show interviewing, speed dating …
• For preserving rare abilities or dying-out knowledge, such as making specialised classifications, authenticating art, identifying banknote forgeries…
26
Knowledge elicitation generally refers to finding out information from a human informant.
Knowledge elicitation: the process of discovering knowledge from a human source, commonly using methods of
observation, interview, questioning and verbal or behavioural analyses.
Requirements analysis: involves investigating a problem situation to identify what the information needs
and required processes actually are, before a solution is designed
KE techniques are often used in developing information systems, to find out what people need and want from the system
27
Some knowledge elicitation methods
• Interviewing
• Protocol analysis
• Prototyping and storyboarding
• Task analysis
• Repertory grid
• Card sorting
• 20 Questions
28
Interviewing• A formalised conversation between two or more
people-One party asks the questions, the other responds
-‘Teachback’ is feedback from the questioner to confirm their understanding of the answer
• Aim is of finding out what the person knows or thinks
• Job interviews, market research, TV chat shows, systems requirements analysis, witness interviews…
• Provide a sense of issues, vocabulary, attitudes of interviewee
29
Structured and unstructured interviews• Structured
-Questions are planned in advance and same questions are always asked
-Easy to compare interviews-But may miss nuances of opinion or detail
• Unstructured-Format is looser and follows where the conversation leads-Can elicit richer information-But can go “off topic”
• Semi-structured-General framework of topics is planned, but interview is managed conversationally
30
Protocol analysis• Helpful in eliciting the procedures that experts use to
solve problems
• Verbal or behavioural:-Behaviour is observed and recorded as an expert works through a problem. Verbally they can talk the elicitor through it
-Protocol is recorded, transcribed and analyzed
• Advantage - knowledge can be captured that the expert perhaps has not, or cannot verbalise. A model of the expert’s knowledge can be created.
• Disadvantage – talking can distort what is going on
31
Task analysis
Task analysis can be applied before protocol analysis-Break task down into stages and required actions
-Underlying structure of tasks and procedural knowledge requirements can be determined
-Results of task processes can be predicted
-“Throwaway” comments by experts in the execution of tasks can provide important insights into procedures.
32
Pick up tooth brush Wet brush Take the cap off tube Put paste on brush Brush outside of the bottom row of teeth Brush outside of the top row of teeth Brush biting surface of the top row of teeth Brush biting surface of the bottom row of teeth Brush inside surface of the bottom row of teeth Brush inside surface of the top row of teeth Spit Rinse brush Replace brush in the holder Grasp cup Fill cup with water Rinse teeth with water Spit Replace cup in holder Wipe mouth on sleeve Screw cap back on tube Place tube back in friend’s toiletry kit so she doesn't realize that you forgot to bring toothpaste on the trip
Task analysis: Brushing teeth
Example from Tom MacIntyre at www.behavioradvisor.com
33
Psychological techniques
• Get past what people say to the meanings by which they construct and understand their world
• How ideas or concepts fit together
-Card sorting
-Repertory grid
34
Card sorting
• Five cards represent my knowledge about Australia
• How to group them?
35
Card sortingAustralia is a country, so I’ll group it above
the citiesNewcastle and Sydney are
both in NSW, so I’ll group
them together
Perth and Brisbane are
both sunny, so I’ll put them in another group36
Card sorting
• Elicits information about how someone organises and categorises ideas
-The reasons for the classification is as important as the categories themselves for finding out how the person thinks
• Used in market research, web page design
• Easy, cheap, flexible
• Also use pictures, sounds, etc
37
Repertory grid• From Kelly’s Personal Construct Theory
-Originally used in psychology, now in many other situations
• Elicits the framework of understanding that a person brings to make sense of their world
• Takes items of interest in a problem situation and aims to identify how an individual thinks about them, using different constructs
• Each construct has two poles that are opposite
-“Sweet defines sour”
38
Rep grid example – classifying fruit
apple pear grape banana peach lemon lime orange
39
After third ranking:
Round (1) Citric (1) Single fruit
(1) apple 1 4 1 pear 3 5 1 grape 1 4 5 banana 5 5 4 peach 1 4 1 lemon 2 1 1 lime 2 1 1 orange 1 1 1 Long (5) Sweet (5) Bunches (5)
40
Representing rep grid results
RED (1)Goes with salmon (1)
Shiraz 1 5Merlot 1 4Pinot noir 1 1Chardonnay 5 2Champagne 5 1Liebfraumilch 5 3Lambrusco 1 3
WHITE (5)Better
with meat (5)
If (wine is red) and (goes with salmon) then (pinot noir)
Wine colour?
Goes with salmon?
Shiraz Pinot noir
red
white
no yes
41
Discovery informatics
Cap colour Edibility
1 Red Poisonous2 Red Poisonous3 Beige Edible4 Beige Edible5 Red Poisonous6 Beige Edible7 Red Poisonous
If red then poisonous(what type of reasoning is this?)
the study and practice of employing the full spectrum of computing and
analytical science and technology to the singular pursuit of discovering new
information by identifying and validating patterns in data.
(Agresti)
42
Knowledge elicitation techniques are required when finding out knowledge that
is as yet unrepresented from human experts.
These techniques can also elicit information about how an expert views and constructs their understanding of a
particular area
Recap
43
Other aspects of discovery
• Inquiry, discovery and meaning making are common to informatics practice and to everyday activities in other walks of life.
• Verifying information by checking different, and preferably original, sources is good practice.
• Discovery may be obtrusive (such as questioning), or unobtrusive (such as automatically-discovered patterns in selling)
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation and
knowledge7. Summary
44
Message-bearing objects
• Another crucial idea in search, research and discovery is the idea that any artefact or document bears a message, as these are representations of intentional human thought or activity
45
Representation
• Representation is the second part of knowledge discovery
• Representation allows what has been elicited to be recorded, stored, shared and generally used.
• The representation is chosen with a view towards this, and to the forms that those receiving the information can interpret.
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation and
knowledge7. Summary
46
Representation
• Describing the essential qualities of representational signs and symbols is the concern of the information disciplines
- Must understand the signs and symbols of a discipline or field, and what they mean at different levels
• In semiotics (chapter 2) different levels of representation apply, from coded data to applied human knowledge
47
8
8• Form
-8, the concept of eightness• Meaning
-8, the house number• Usage
-where the party is• Understanding
-between houses 6 and 108 88 8
848
A representation “of” something
• Representations in the middle are called mediating representations
Is a representation “to” someone
49
Mediating representations
• Representations between a source of knowledge and a target
• Bridge the communication gap between the verbal data coming from the source and an operational form oriented towards future computation
• Highlight what has been discovered in ways that both see as an expression of that idea.
50
Intermediate representation
• A model of the knowledge area that integrates all the specific mediating representations
- A ‘complete description’ of the topic, using rules, trees, grids, diagrams and structured English descriptions, glossaries…
• Can then be translated via other intermediate representations into a computer program
- At this point the human source is out of the picture, and the mappings between representations are formally defined.
51
Representation in informatics
• Different fields of informatics use different types of representational forms, specialised for their subject:
-Business informatics may use charts and line graphs-Systems designers may use data flow diagrams or rich pictures
-Social informatics analysts may use directed graphs to show social connections
-Computer programmers may transform other representations into languages and data structures
- Bioinformatics may use visualisation tools to explore large data sets
-….52
Categories of representation
• Graphics - Transform data sets into visual equivalent
• Sets and logic- Enable reasoning about data
• Modelling- Represent a selection of information and embody an interpretation of it
-Causal loop diagrams- Influence diagrams-Conceptual mapping
53
Charts
54
Chernoff faces
• Facial features are used to represent different aspects of the data
• Humans are very good at recognising faces - so can easily notice differences among faces, which indicate differences in the underlying data patterns
Image courtesy Bradley Mohr 55
56
Discrete mathematics
• The branch of maths dealing with sets of elements• Some basic ideas:
-Sets-Logic-Combinatorics-Functions-Graph theory-Probability
• We will meet many of these ideas in this chapter and elsewhere in the book
57
Sets
• A set is any well-defined collection of elements- The order of elements does not matter- Each element occurs only once
• Venn diagrams are used to represent sets pictorially• Some set concepts: universal set, empty set, subset,
intersection, union, difference
58
• The universal set U is the set of all PETS• Dogs and Cats are subsets of the universal set• The set of Dogs and the set of Cats have no animals
in common59
• The set of Lapdogs is a subset of the set of Dogs
60
• The set of my pets intersects with the sets Dogs and Cats (I own a dog and a cat)
61
Logic
• Sets represent propositions, or statements that can be true or false:
All lapdogs are dogsNo dogs are cats
• Logics enable propositions to be combined in a syllogism and deductions made:
> No lapdogs are cats
62
Representing logic in Prolog
FACT: Fido is a dog
PROPOSITION: All dogs are smelly
which is the same thing as “if X is a dog then X is smelly”
dog (Fido)smelly(X):- dog (X)
Replace X with Fido to deduce the new fact that Fido is smelly. 63
Causal loop diagrams
• Show causal influences, or how someone thinks things they observe in the world fit together
“windy conditions causes waving tree branches”.
(or perhaps waving tree branches cause windy conditions?)64
Causal loop diagrams
65
Influence diagrams
• Also show beliefs about causes and relationships, but also allow probability information to be included for decision-making and prediction
• Also called decision networks or belief networks• Show how ideas are related, along with the strength
with which they influence one another
66
It might rain: should I take an umbrella?
Influence diagram67
Maps
• Maps lie on a range from simple descriptions with subjectively meaningful associations (such as in rich pictures), through to detailed models.
• Maps may represent:- an externally organised reality
-a road map corresponds to physical roads in the outside world.-or a convenient and communicable impression of some constructed understanding
- this book provides a map to guide you through the territory of informatics.
68
Maps
• All maps have a structure in which concepts are associated by relationships in some meaningful way
• These concepts and relationships have different names in different fields of study
-Entities and relationships-Nodes and arcs-Vertices and edges-…
• Formal map structures may include graph structures with mathematical properties, such as directional links (as in causal loops) and numerically weighted links
69
Rich pictures
• Represent the problem elements in a situation as the individual stakeholders see them
• Originally part of Checkland’s Soft Systems Methodology (SSM)
• Include:-Structural aspects of the situation-Processes and flows-Stakeholder concerns
• Reflect values, emotions, attitudes – political and social context of problem
70
A tavern as a rich picture
71
Conceptual mapping
• Diagramming technique used to visualise relationships among concepts
• Mind map• Concept map• Sociogram
72
Mind map
• Originated by Tony Buzan• Visual overview of an area• Central topic with branches and sub-branches indicating
relationships• Colours, shapes, symbols are chosen to be personally
memorable
73
Mind map of Informatics
74
Concept mapping
• Originated by JD Novak • Based on associative theory of human memory• Concepts are organised hierarchically, with relationships
named to connect concepts into meaningful sentences
75
Concept map of Informatics
76
… what does informatics mean to you?
Draw your own mind map or concept map!
77
ValDiarmuid
John
Tanya
Paula
Pete
John and Val have published a lot together
John and Pete have written together …
… as have John and Paula
… and so have Val and Diarmuid.
… Val and Tanya.
So those lines are drawn thicker.
Sociogram showing network of publication patterns within a group
78
Some common representation techniques in informatics
include graphs, sets and logic, and mapping and modelling
Recap
79
Discovery, representation and knowledge
• As representations are at least one step removed from their source they take on an existence of their own.
• They reflect different states of knowledge: both a reduction from the originating source, and a “complete resource” providing everything now available to be interpreted by the receiver.
1. Introduction2. Discovery3. Knowledge elicitation and
discovery4. Other aspects of discovery5. Representation 6. Discovery, representation
and knowledge7. Summary
80
Discovery, representation and knowledge• Representations present a symbolic experience to their
users, who, as the bringer of meaning to the event, may freely select, misinterpret or misunderstand what was intended by the source of knowledge.
81
Checking understanding
• If representation is to be able to transfer information across time and space, the source and the recipient must share the same interpretation of it
• When producing any representation, check it against the original source to ensure the source and recipient share the same interpretation
82
Meaning is given to a representation in the context of interpretation.
Checking understanding against a knowledge source is therefore a key process to apply when producing any
representation.
Recap
83
Chapter summary
• Representation allows what has been discovered to be recorded, stored, shared and generally used
• Searching techniques are used when represented knowledge already exists
• Knowledge elicitation techniques are needed when knowledge is still “in people’s heads”
• Different disciplines have their own conventions for representations
• Shared interpretation of a representation is essential to ensure shared meaning
84
85