Crafting a Research Paper/Talk

Some initial principles/patterns
Prasun Dewan
SN 150, Sitterson, 11-12:15
962 1823
[email protected]
Software vs. Communication
• Correctness vs. Style
• Style helps correctness
• Style more often abused
Deriving Principles/Patterns
Reusability is good
Cost of re-using
software is lower than
writing new software
Encapsulate as client of
object does not react to
implementation changes
Use MVC as view can be
changed without
changing model
• Start with axioms
– Defend but not prove
them
– Often considered goals
– Should be as few as
possible
• Every principle/pattern
should not be an axiom!
• Derive
principles/patterns
from them
Axioms/Goals of Research Talk/Paper
• Understandability
– Better not communicate than be unclear
• Novelty (Comparison with related work)
– Not considered research otherwise
– People need to be convinced to some extent work is novel
• Cleverness
– Tedious work not considered research
• Work amount
– Otherwise contribution not significant
• Attention
– First few minutes crucial
Other metrics?
Talk vs. Research
• Novelty
– In research
– Shown in talk
• Cleverness in
– Research
– Shown in talk
– In composing talk
• Work amount
– In research
– In talk
Work amount and cleverness in talk are secondary and important goals
Relationship
Under.
Novelty
Cleverness
Work. Amt.
Attention
Under.
Novelty
Cleverness
Work. Amt.
Attention
• Many goals conflict with each other!
– That is mainly why talks are hard even for experienced
presenters
• Some support each other
Understandability
Under.
Under.
Novelty

Cleverness
Work. Amt.
Attention

Novelty
Cleverness
Work. Amt.
Attention
• Loss of attention when not understandable
• Without clarity, novelty hard to determine
Novelty (Comparing with Related work)
Under.
Cleverness

Under.
Novelty
Novelty

Work. Amt.
Attention



Cleverness
Work. Amt.
Attention
• Not effectively distinguishing related work may make contribution
seem less clever
• People may pay more attention if they know others have worked on
subject area
• Bringing out the relationship with something reader knows may
improve understanding
Cleverness
Under.
Novelty
Cleverness
Work. Amt.
Attention

Under.
Novelty

Cleverness





Work. Amt.
Attention
• People like to listen to insightful talks
• People realize that clever things take effort
– Amount of effort depends on person
– A smarter person might have larger expectations!
• Clever things are hard to explain

Work Amt.
Under.
Novelty
Cleverness
Work. Amt.

Under.
Novelty

Cleverness

Work. Amt.

Attention






Attention
• Adding anything to talk increases chance of
something not being understandable (assuming
same amount of time)
• May go less deep and thus not show cleverness
Attention
Under.
Cleverness
Work. Amt.
Attention

Under.
Novelty

Cleverness

Work. Amt.


Attention
Novelty








• If people are not paying attention, all is lost!
• Assuming attention is on relevant material

In remaining course
Under.
Novelty
Cleverness
Work. Amt.
Attention
• Techniques (principle/pattern) with specific
examples and arguments based on
axioms/goals
• Will analyze
– common rules of thumb (many are on the web)
– identify our own
Patterns and principles
• Some initial patterns and principles to seed
the process
• Judge against your own rules and talks
• Think of how principles implemented
– Specially flow
• Think of what you could add to flow
Identify Potential Applications?
Understandability
Novelty
Cleverness
Work. Amt.

Attention

• E.g. Talk/paper patterns should improve talk quality
• Improves attention as people like practical results
• Reduces time for work amt. but is important if audience
not familiar with applications
• Not a goal as some research may not have applications
• Don’t make talk too application-centric
– Idea-centric
Related Work?
Understandability
Novelty
Cleverness


Work. Amt.
Attention

• Needed to prove novelty
• Contribution obvious only in retrospective
– Related work shows “wrong” way to do it
• Supports mystery story and thus attention
– After a way that does not work, people want to know a
way that does
Compare as sets of features?
• IR Control Programs PocketPC/Palm
• Cooltown – HP (2003)
• MOCA – IBM (1999)
• Universal Plug and Play –
Microsoft (2003)
• Jini (Service UI)– Sun (2001)
• Personal Universal
Controller (PUC) – CMU
(2004)
• Hodes’ System – UC
Berkeley (1998)
• ICrafter – Stanford (2003)
• List all systems
• Say our system has new set
of features
Say Something Intuitive
• Hodes’ System – UC
Berkeley (1998)
• Our infrastructure looks
at user centered
whereas theirs is
system-oriented
Show Holes in Design Space
UI Deployment
UI Generation
Client
Fully
Automatic
Predefined (UI)
Remote
SemiAutomatic
Fully
Automatic
Client
Factory
SemiAutomatic
Design space if often a
contribution in its own right
Remote
Factory
Device
Factory
3rd Party
Factory
Identify Evaluation Space
UI Deployment
UI Generation
Client
Fully
Automatic
Predefined (UI)
Remote
SemiAutomatic
Fully
Automatic
Client
Factory
SemiAutomatic
Design space if often a
contribution in its own right
Remote
Factory
Device
Factory
3rd Party
Factory
Classify related Systems
Client Factory
Predefined
Approach
Device
Factory
Universal Plug and Play
MOCA
3rd Party
Factory
UI Generation
Palm/Pocket PC IR Control
Programs
Cooltown
Client-side
Remote
Jini (Service UI Approach)
Personal Universal Controller
Hodes’ System
ObjectEditor
ICrafter
Identify Evaluation Metrics
• User-Interface Flexibility (Qualitative)
– range of user-interfaces an approach can support
• Programming Costs (Qualitative and Quantitative)
– amount of code required to deploy a user-interface
• Maintenance Costs (Qualitative)
– programming time and resources required to support and
update user-interface code
• Efficiency (Qualitative and Quantitative)
– time and storage space costs of an approach
• Device Binding Time (Qualitative)
– time a client must learn about (or bind to) a device in
order to deploy a user-interface for it.
• Deployment Reliability (Qualitative)
–
the level of guarantee an approach offers in deploying a
user-interface
Often contribution is some new set of metrics
Compare With Related Work
• Be Sure to Point Out Advantages and Disadvantages
• Can give results without proof in conference talk
– But don’t shy away from complexity in longer talks
New Result Research?
• Must have some new “idea”
– Retarget user-interface for device of one kind to userinterface for another kind of device so common parts are not
re-created
• Ideally should compare with related ideas in even
different domains
– Caching
New IdeaResearch?
• Must show there is complexity
• Various Ways
– Equations, Architecture,
Abstraction, Algorithm
• Do not need to give all details
– But do not shy from complexity
in job and thesis talks
– Should describe at least one
component in some depth
Practice Makes Perfect
Understandability
Novelty
Cleverness
Work. Amt.




Attention

• Can show more content
– TV news, ads convey so much information per unit of time
– Both work amt and novelty
• Uh, ah, “you know”, pauses, groping for words, lack of
confidence reduces attention
– Many good speakers talk as fast as they can without mumbling!
– Lampson units of speed.
• Can improve ways to make material understandable and look
clever
– Assuming iteration
• Nature, cons, factors on which it depends?
Nature of Practice
•
•
Rehearse in your mind (until the last
moment)
Speak it out in front of the mirror and
record audio
–
•
•
Use recording software to record slides &
audio
Rehearse in front of one person (e.g.
advisor/co-author)
–
–
•
Next option is perhaps a better alternative
May not feel as much energy as with an
actual audience
Can use it in addition or in place of next
option.
Rehearse in front of a practice audience
–
Most important if you can get such an
audience (record it!)
Memorize the speech?
• Can look too rehearsed
• Realty TV better than someone
delivering an obviously
memorized script?
• But movies, plays are rehearsed,
learn to be an actor!
– Amazing similarity in same talk given
at diff. times by great speakers
– Need to put appropriate pauses
• Rehearsed drama delivered more
or less naturally better than realty
TV
Memorization Technique
• Problem can occur if you start
with script before talk
• Create script/memorize of what
worked after each iteration!
Nature of Script
• Must keep conversational style
– Point to screen rather than describe
– At least in computer science
– In softer fields often writing/oratorial
skills demonstrated in talks and
speeches are often read
• Often in a very verbose way
• Talks by non-native speakers typically
have more content!
Example: Conversational Style?
• The design pattern does not
define if the model and editor,
which, recall, performs input
and output, are centralized or
replicated. So let us consider
these architectural issues.
• You might as well be reading
paper.
Example: Conversational Style
• The design pattern does not
define if the model and editor are
– Centralized
– Replicated
Centralized
• So let us consider these
architectural issues.
• Graphics and animation improves
conversational style
– Words explain graphics on screen
– Like slide show
Replicated
Cons of SlideShow Approach?
• PowerPoint takes center stage
– Many think of PPT as a bad thing
– In business not research
• Graphics for abstract ideas a la abstract
art
• Much harder as not reading text
– Can animate text points in case cant
remember
– Must use grammatically correct(and
ideally elegant) English to expand points
• Otherwise better to read text slides
– Need script and practice to really make it
work (next slide is example)
Product
Problem
• Brooks ‘74:
Diminishing returns
as people are added
to project
• Many reasons
–…
– Conflicts
Way back, Brooks found that t(click) adding people to a project does not
result in proportional increase in team productivity. Over the years, people
have found many reasons for this problem. One of these is conflicts, not
among people, as in the talk before, but between the code they write in
parallel software development activities.
Amount of Practice
How much to practice?
• Con: takes time and is
tedious!
• Inversely prop. to time
available for talk
– I did not have time to
write a shorter letter
– Cannot afford pauses
• Inversely prop. to how
articulate/experienced
• Directly proportional to
importance
Practice in Different Kinds of Talks
• Defense
Amount of Practice
C
C
D
– Consequence can be failure
– Committee knows work
– Job talk precedes
J
J
• Conference talk
D
L
• Job talk
C
L
– Shortest possible time
– Potential interview
J
– Decides future!
– Conference talk precedes
D
L
• Class lecture
– Cannot afford overhead
– Audience asks clarification
– Job talk precedes
Don’t Hide Information in Slide
• Belief strongly held by many
– Covering ports of transparency
considered bad and distracting
• Understandability
• Belief strongly held by many
– Covering ports of transparency
considered bad and distracting
– Audience member can go
ahead of speaker and get more
context
– Audience member can go
ahead of speaker and more
• Like animation in slide
context
– Graphics hard to understand all
– Figures hard to understand all
at once
at once (Satyajeet example)
– Animation in slide
– Can indeed provide mystery
– Cannot hide answer to
– Cannot hide answer to
question
question
Animation
Understandability


Novelty
Cleverness
Work. Amt.

Attention

•
•
•
•
Helps incremental understanding
Keep attention as reader may go ahead
Useful in Socratic (question and answer) explanation
Too much animation can prevent understanding as less
context available at any time
• Animation takes time
– Future work may not be animated
• Consider two alternatives shown in next two slides
MODELING MULTI-USER INTERACTION
Application
Coupling
Active Display
Active Display
Conflict Management
Editing
Commands
User 1
Undo/Redo
Editing
Commands
User 2
MODELING MULTI-USER INTERACTION
Application
Coupling
Active Display
Active Display
Conflict Management
Editing
Commands
User 1
Undo/Redo
Editing
Commands
User 2
What should be animated?
• Parts that need to be grasped incrementally
• An answer to a question
• ….
Types of Questions
IQ
EA
• E.g:
EQ
– Should (parts of) a talk be structured as
a series of questions and answers?
RQ •
RA
IA
Explicit
– Audience given chance to answer
– Expected to provide simplified answer
• Rhetorical
– Answered by presenter
• Implicit
– Raised in audience mind as side-effect
of some information given by
presenter
– answered in subsequent slides
Question and Answers?
Understandability

Novelty
Cleverness

Work. Amt.
Attention

• All
– Audience pays attention
• They try to answer question
• Connection is made to audience, so they will more friendly, and
thus more guilty about going to laptop
– Favors understandability if audience tries to answer
– Favors cleverness if audience is thinking of wrong answer
– Specially if in retrospect, answer is simple, as good
solutions should be
Question and Answers?
Understandability
Novelty
Cleverness
Work. Amt.



• Explicit and rhetorical
– Conflicts with cleverness
if audience easily thinks
of correct answer

Attention

• Explicit
– Unsettling if audience does
not attempt answer
– Survey questions are safest
• How many of you use the
model-editor design pattern
– Explicit questions can be used
to adapt talk?
• Do I need to explain modeleditor version
• People will not say yes lest
that will end up boring others
• People know less than you
think!
– Takes time, works in lectures
Question and Answers?
Understandability

Novelty
Cleverness
Work. Amt.

• Implicit
– Implies audience paying attention
– None of the above disadvantages
– Makes talk like a mystery story
Attention

Make Talk a Story?
Understandability
Novelty
Cleverness
Work. Amt.
Attention

• Creating flow – connecting each information
item to (ideally immediately) preceding item.
• Favors attention
– Even TV/radio news writers try to create a flow!
Make Talk a Story?
Understandability
Novelty
Cleverness

Work. Amt.
Attention

• Favors cleverness
– talk looks like proof with subsequent steps following
from previous ones
– work looks like one big contribution than collection
of unrelated small details
– thought that went into talk appreciated
Make Talk a Story?
Understandability
Novelty
Cleverness



Work. Amt.
• Favors novelty if previous work part of story
• Favors understandability as people see the a
coherent picture
Attention

Is Creating a Flow Hard?
• Explicit and rhetorical
– Audience pays attention
– Conflicts with cleverness if
audience easily thinks of
correct answer
• Implicit
– None of the above
disadvantages
– Makes talk like a mystery
story
Ordering used for –2 slide
• E.g. -1 slide  this slide: “ls
creating a flow hard?”
• E.g. -2 slide  “this is a
special case of the more
general rule that a talk
should be a story”
• First example relatively easy
• Second example required
special ordering
Is Creating a Flow Hard?
• Implicit
– None of the above
disadvantages
– Makes talk like a mystery
story
• Explicit and rhetorical
– Audience pays attention
– Conflicts with cleverness if
audience easily thinks of
correct answer
Alternative equally good ordering for intra -2
slide flow, but not inter-slide flow
• E.g. -1 slide  this slide: “ls
creating a flow hard?”
• E.g. -2 slide  “this is a
special case of the more
general rule that a talk
should be a story”
• First example relatively easy
– But must remember transition
• Second example required
special ordering
Indicators of bad flow
• Simply stating the slide
title
– Even paraphrasing is not
enough
• “ I will next talk about
…”
• “Another component of
the approach is …”
Indicators of good flow
• Connection to previous
slide
– “The concept leaves
several questions …”
– “This idea has the
problem/advantage …”
– “A related issue is …”
• Connection to far away
slide
– “The story so far is …”
“One issue I have not
addressed is”
– “I will connect this to …
later” “I had mentioned
that I would find a
connection to … This
slide does so.
Special Slide for Flow
• “The design framework does not
define if the model and editor is
– Centralized
– Replicated.
• So let us consider these
architectural issues.”
Outline/Road-Map?
• Design Pattern
• Architecture
• Outline can create flow
– Users have been told the
sequence of items
– Can bring back outline to
go to next topic in it.
Mediocre Outline Based Flow
• Design Pattern
• Architecture
• “I will next talk about
the architecture”
• “Another part of this
work is the
architecture“
Better Outline-based Flow
• Design Pattern
• Architecture
• The design pattern does
not define if the model
and editor are
centralized or
replicated. So let us
consider these
architectural issues
Build Outline Incrementally
• Design Pattern
• Architecture
• Probably said the same
thing when first
showing outline
• So better to lose the
outline at start
• May build it
incrementally
Main Outline?
Understandability
Novelty
Cleverness
Work. Amt.
Attention

•
•
•
•
•
•
Problem
Related Work
Approach
Evaluation
Conclusion
Future Work
• Most talks have similar main
outline
– Main outline in stories?
• First few slides decide if people
will pay attention
• How to start the talk?
• Consider two examples
Semi-Synchronous Conflict Detection
and Resolution in Asynchronous
Software Development
Prasun Dewan
Rajesh Hegde
University of North Carolina
Microsoft Research
[email protected]
[email protected]
The problem we are solving has to do
with collaborative software development
Product
Problem
• Brooks ‘74:
Diminishing returns
as people are added
to project
• Many reasons
–…
– Conflicts
Crafting a Research Paper/Talk
Prasun Dewan
SN 150, Sitterson, 11-12:15
962 1823
[email protected]
Paper/Talk
• Paper: document
created
• Talk
– Slides and/or
Delivery
– Some talks do not
have slides!
– Talk = LiveMeeting
Recording
Crafting
The passive voice should
not be used!
Use light text on dark
background !
Have an abstract,
introduction, body ,
conclusions and future
work
Have a title, outline, body,
conclusions , and future
work
• In the small
– Grammar, PPT Animations
– Style, PPT Color Choices
– Analogous to defining an object
• In the large
– Composition of prose and slide
items
– Analogous to design patterns
• Assume proficiency in design in
the small
State of the art in Papers/Talks
• Arguably good
composition
techniques
• Situational
• Examples!
• Practice
No one seems to have looked for patterns!
Hypothesis: Such Patterns Exist
Each student seems to make the same kinds of mistakes!
How to start the talk?
Crafting: Composition of
prose and slide items
Talk: Slides + Delivery
Patterns; Arguably good
general compositional
techniques.
No one seems to have looked
for patterns!
Axioms
• Define one or more terms of title
– If necessary
– If not, do not read the title, or text on
any slide!
• Motivate
– If necessary
• Give unsolved problem raised by
subject of talk
– Describe state of art
– In research, not industry
• Start describing solution
– Everything else should be connectable
to problem and solution
Inductive Flow
Crafting: Composition of
prose and slide items
No one seems to have looked
for patterns!
• Define terms of title
– If necessary
• Give unsolved problem raised
by subject of talk
– Describe state of art
– In research, not industry
• Start describing solution
Axioms
– Everything else should be
connectable to problem and
solution
Deductive Flow
Crafting: Composition of
prose and slide items
No one seems to have looked
for patterns!
• Define terms of title
– If necessary
• Give unsolved problem raised
by subject of talk
– Describe state of art
– In research, not industry
• Start describing solution
Axioms
– Everything else should be
connectable to problem and
solution
Deductive vs. Inductive
Abstraction
• Inductive learners like to
work out general principles
from cases and examples
– Retain information better
– Attempt to solve mystery
Cases /examples
Abstraction
Cases /examples
• Deductive learners like to
see general principles and
then cases and examples
– Can become good scientists
– Happier with non mystery
Inductive?
You flirted with my
boyfriend
You were rude to my
mother
I HATE YOU!
You kicked my cat
Example taken from Wolfgang Gatterbauer
Deductive?
You flirted with my
boyfriend
You were rude to my
mother
I HATE YOU!
You kicked my cat
Example taken from Wolfgang Gatterbauer
Alternative Deductive vs. Inductive
• Inductive (Supp. Args.)
Conclusion
– May not clear be why information
is being given
– Too much mystery!
• Investigation described without
identifying crime
• Deductive (Supp. Args)
Supporting Arguments
Conclusion
Supporting Arguments
– Point is clear
– Creates flow/story
• Motivated vs. unmotivated better
name
• Will implicitly assume this
deductive, called the Minto
pyramid principle
• Not to be confused with
(information pyramid (later)
Illustrate with Examples?
Understandability

Novelty
Cleverness
Work. Amt.

Attention

• For related work and
bringing out
requirements
• Helps understandability
– Needed for inductive
– Helps deductive
• Keeps attention
– Specially if example is
real-world
Illustrate with Examples?
Understandability

Novelty
Cleverness
In the small
Grammar, PPT Animations
Style, PPT Color Choices
Analogous to defining an object
In the large
Composition of prose and slide items
Analogous to design patterns
The passive voice should
not be used!
Have an abstract,
introduction, body ,
conclusions and future
work
Work. Amt.

Attention

• In soft fields, field without
benchmarks, or talk or
conference paper
– definition/proof by example
important to make point
• Can take time
• Use running example to
– reduce time
– bring out all issues for
inductive
• Next few slides are examples
Example Conflict
public Shape (int initWidth, int initHeight ) {
…
}
Alice
Refactors to change parameter order
public Shape (int initHeight, int initWidth) {
…;
}
Changing APIs: de
Souza, Redmilles et al
‘04
Subclasses with old parameter order
Bob
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Traditional Conflict Management Model
public Shape (int initWidth, int initHeight ) {
…
}
Bob has not checked in as yet
Check-In
Traditional Conflict Management Model
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Check-In
Diff
• Asynchronous Software
Development
• Line-based Conflict Detection
• Individual Conflict
Management
• Late Conflict Management
Merge
Compile
Test
New Requirements and Model
public Shape (int initWidth, int initHeight ) {
…
}
New Model
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
• Asynchronous Software
Development
• Dependency-based Conflict
Detection
• Collaborative Conflict
Management
• Early Conflict Management CollabVS = Visual Studio + Semi
Synchronous Collaborative Conflict
Management
Incremental Dependency Checking
public Shape (int initWidth, int initHeight ) {
…
}
Make Next Edit
Calls
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Make Next Edit
False Positives
Cannot be
Eliminated:
Halting Problem
Conflict Inbox
View Next
Warning
Switch to Detailed
Conflict Warnings
Make Next Edit
Dependency
Notification
• Email metaphor
Switching to Non-Conflicting Work
View
Next
Next
SetView
Watch
Warning
Warning
Switch to Edit
Context
Make Next Edit
public void NonConflicting () {
}
Conflict Prevention
View
Next
Next
SetView
Watch
Warning
Warning
public ARectangle(int initWidth, int initHeight):
base (initHeight, initWeight){
….
}
Switch to Edit
Context
Make Next Edit
Watch
Notification
Switch to
Refreshable Code
Session
• Can change parameter order
Interesting vs. Crucial Examples
Rehearse in your mind (until
the last moment)
• Interesting example simply
provides prop for script
– Often picture worth a
thousand words
• Increase attention
– Practice to see if they are
too frivolous
• Don’t need to use any
words for them and hence
take no time
Graphics vs. Non-Graphics?
•
•
•
•
Images
Architecture
Icons
Anything non-bullet?
Interpreting graphics
• Look at audience
– You are the focus
• Audience listens to you
– May not notice graphics
– Specially fast moving animation
• Look at slide
– Audience looks at you and slide
• Do not look into space
Referencing a screen area
• Pointing devices
– Distracting, becomes focus of attention
– Always usage issues
• People circle rather than point
• Audience has no idea
– Mouse, laser position sometimes hard to see
• Direct pointing
– Makes you move
• Do not just stand at podium like a statue
– May not be possible in really big conferences
• Animate object on which you want to focus
– Animation could be missed
Illustration Order
• Illustrate incrementally
Incremental Dependency Checking
public Shape (int initWidth, int initHeight ) {
…
}
Make Next Edit
Calls
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Make Next Edit
False Positives
Cannot be
Eliminated:
Halting Problem
Conflict Inbox
View Next
Warning
Switch to Detailed
Conflict Warnings
Make Next Edit
Dependency
Notification
• Email metaphor
Alternative Order
• Give complete model
• Then illustrate
Complete Model
Incremental Dependency Checking
public Shape (int initWidth, int initHeight ) {
…
}
Calls
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Conflict Inbox
• Email metaphor
Alternative Order
• Give complete model
• Then illustrate
• Deductive by definition
Yet another alternative
• Illustrate
• Then give complete
model
• Inductive
Incremental Dependency Checking
public Shape (int initWidth, int initHeight ) {
…
}
Calls
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Conflict Inbox
• Email metaphor
Complete Model
Lazy “Evaluation”
Understandability

Novelty
Cleverness
Work. Amt.
Attention

• In lazy evaluation, expression evaluated just before use
– If not needed, not referenced
• Provide information incrementally
• Or define just before use (and not much earlier or later)
• Or do not wait too long to motivate or illustrate some
concept.
• Or do not provide information irrelevant to conclusion
• Otherwise will repeat or will lose people
• In deductive will not motivate
• Judgment call as to what is atomic unit of information
Repetition?
Understandability
Novelty
Cleverness
Work. Amt.

People learn through
repetition
People learn through
repetition
Attention

• People learn through repetition
• Quick learners can get bored
• Conflicts with work. amt.
– May not be able to explain
something else
• How many times?
• Tell what you are going to tell them, Tell
them, Tell them what you told them
Summary-Detail-Summary
Repetition should be
structured into three parts
Tell them what your are
going to tell them, Tell
them, Tell them what you
told them
To summarize, repetition
should be structured into
three parts
• Abstract the topic
• Give next level(s) of
detail
– Information pyramid
• Summarize the topic
– Perhaps later in the
conclusion section
when people have
forgotten
Topic-Detail-Summary
Let us consider the nature
of repetition
Tell them what your are
going to tell them, Tell
them, Tell them what you
told them
To summarize, repetition
should be structured into
three parts
• Identify the next
topic
• Explain the topic
• Summarize the
topic
Repetition for Non-Linear Flow
As we will see later,
repetition can be useful.
Some other topic.
Tell them what you are
going to tell them, Tell
them, Tell them what you
told them.
Some other topic.
As I mentioned before,
some form of repetition
may be useful.
• A talk/paper often cannot
be a linear sequence
– Tree, Hyper-graph
– Try to create linear flow!
• Repetition for forward
referencing
• Repetition for reminding
• Be sure to point out that
you are repeating
– Otherwise people get a
feeling of déjà vu and tune
out
Repetition for Giving Full Picture
• Present parts of the
model
• Then put it all together
– If there is time
Incremental Dependency Checking
public Shape (int initWidth, int initHeight ) {
…
}
Make Next Edit
Calls
public Rectangle(int initHeight, int initWidth):
base (initHeight, initWeight){
….
}
Make Next Edit
False Positives
Cannot be
Eliminated:
Halting Problem
Conflict Inbox
View Next
Warning
Switch to Detailed
Conflict Warnings
Make Next Edit
Dependency
Notification
• Email metaphor
Complete Model
Next two slides are inductive slides
Theoretical Evaluation
• Asynchronous Software
Development
• Early conflict detection
• Dependency-based
conflict detection
• Collaborative conflict
detection and
resolution
That was our theoretical
evaluation. Next Rajesh will
describe the lab study we did.
Lab Study
• 16 developers, Groups of 2 (A & B)– not colocated (A and B did different tasks)
• Training -20 minutes
• Actual task -60 minutes
• Survey, Debrief – 15 minutes
Theoretical Evaluation
• Asynchronous Software
Development
• Early conflict detection
• Dependency-based
conflict detection
• Collaborative conflict
detection and
resolution
Thus the model meets all of our requirements. So it seems we have
accomplished our mission. Well not quite. These requirements were derived
from theoretical arguments. To determine if programmers really wanted to
be liberated from the traditional model, we conducted a lab study, which
Rajesh will describe.
Lab Study
• 16 developers, Groups of 2 (A & B)– not colocated (A and B did different tasks)
• Training -20 minutes
• Actual task -60 minutes
• Survey, Debrief – 15 minutes
Make connections explicitly
• Good flow requires explicit connections
• Even experts may not realize connections
– You may not either
• Standard does not mean required
– Argument that paper cannot get accepted
without lab study is a cop-out
Concise
Understandability

Novelty
Cleverness
Work. Amt.

Attention

• Less is more
– Talk should be just the title?
• Use the minimum amount of words required to
make the point
• Slide-deck can be concise but not presentation
– Practice!
• Holds attention
• Increases attention
• May conflict with understandability
– Sometimes alternative ways are needed
Time in Different Kinds of Talks
Time Available
• What to put in each
talk?
Defense/Job
Conference
Information Pyramid
Abstraction
Abstraction
Abstraction
• Give information at
different levels of
abstraction
• News articles can be cut at
any point from the bottom
– So can talk!
News Example (Philip Yaffe)
Britain yesterday has once again called for the United Nations to mount a peacekeeping
operation in the violence-torn Darfur region of Sudan in response to increasing complaints
from aid agencies on site that international efforts to help Darfur's desperate, displaced
population are woefully inadequate.
At the same time, Her Majesty's Government is joining with other European Union
countries to threaten sanctions against Sudan unless its government energetically moves
to end the "ethnic cleansing" against black villagers in Darfur by the mainly Arab Janjawid
militias. UN officials report that the conflict has already claimed from 30,000 - 50,000
lives and about 1.2 million people have been displaced, with about 200,000 taking refuge
in neighbouring Chad.
News Example (Philip Yaffe)
1. The British Government is concerned about the situation in
Darfur.
2. Darfur is a violence-torn region of Sudan.
3. Britain believes a peacekeeping force is urgently needed.
4. It is pressing the United Nations to supply this peacekeeping
force.
5. This is not the first time that it has urged the UN to supply
peacekeeping force.
6. The population of Darfur has been displaced.
7. Aid agencies in Darfur say that international assistance to
these distressed people is inadequate.
News Example (Philip Yaffe)
1. The trouble in Darfur is a race war
2. Arab militias are attacking black villagers.
3. Britain and other EU countries believe the Sudanese
Government is not doing enough to stop the war.
4. They threaten sanctions against Sudan if its government does
not quickly take action to end the attacks.
5. To date, between 30,000 - 50,000 people have been killed.
6. About 1.2 million have been displaced.
7. About 200,000 have fled across the border into the
neighbouring country of Chad.
8. These figures come from the United Nations, which is a reliable
source.
News vs. Research
•
News
–
–
–
•
may not contain analysis.
Goal is to inform
Story
Research
–
–
–
Shows non-obvious result
Needs some suspense!
Mystery story
•
Don’t want to say butler did
it at the start!
Information Pyramid in Research
•
Conference
Defense, Job
How to create information
pyramid?
Application
• Conference
Conference
– May omit if nothing new
in driving problem
– Networked devices vs.
soft real time
constraints in
multimedia
• Defense/job
Defense, Job
– Give standard ones in
your field
– People outside field are
very skeptical
Related Work
•
Conference
–
–
Conference
•
–
–
Defense/ Job
Usually no time to explain others’ work
Can compare with familiar state of
practice rather than research
–
Physical remote controls vs. software
remote controls
Could give evaluation metrics and state
results of evaluation
May give design space and motivate
metrics if these are original
Comparison can be done at end of talk
rather than at beginning specially when
related work is very loosely connected
Related Work
• Defense/Job
Conference
Defense/ Job
– The “obvious“
relevant ones
– But nothing
irrelevant
Ideas
• Conference
Conference
– The fewer that
encompass the work
the better
– Abstraction is key here
• Defense/Job
Defense, Job
– Name and ideally
describe at a high level
all relevant ideas
Complexity
• Conference
Conference
– Flashing complex
picture/diagram/equat
ion?
•
•
Defense, Job
Cheap way of showing
complexity
Works for showing
large amt of data
– Explain equation,
diagram without
proving/justifying
Complexity
• Defense/Job
Conference
Defense, Job
– Provide the
conference-level
intuition
– Go deep into one
aspect of work
Concise vs. Information Pyramid
• Typically no time to
explain at lower/multiple
abstraction levels
• Choose the abstraction(s)
level given constraints
– Candidate day, job fair,
conference talk
– Prepare lower
abstractions even if no
more time
• To answer more detailed
questions
Tailor Talk to Audience?
• Audience knowledge:
– Experts may know details of research context and
related work
– But experts may like seeing how you define it
• If they are evaluating you
– Can almost never assume audience all experts
• Maybe conference talk
– Assume undergrad knowledge from audience
– Don’t ask “do I need to explain this?”
• People don’t volunteer ignorance
• Just explain at the level of abstraction needed to make
your point
Tailor Talk to Audience?
• Audience’s research accomplishments
– Refer to expert audience member’s
results
• if relevant in making your point
• People love being referenced but not
unnecessary flattery
Tailor Talk Style to Audience?
Abstraction
Cases /examples
Abstraction
• Inductive vs. Deductive
– People judging you likely
to be inductive
– These are the people
who typically succeed in
academics
– Though many successful
(computer?) scientists
are deductive
Tailor Talk to Audience?
Always
Inductive
• Fundamentalists
– There are absolute comm. rules
• Situational Theorists
Inductive
Deductive
– Whether and how communication
understood depends reader’s mind
– There are no absolute rules
– Bad communicators become
fundamentalists
• Those who cant do talk
– Those cant teach, teach about talking
Tailor Audience to Talk?
• Choose audience appropriate for
talk.
• Or don’t give a talk if audience is
not appropriate
• Or Have something to say
– That is interesting to the audience!
• Better to tailor talk to audience
– In some places a thesis synopsis has
to be given to parents and friends
and family
– Build information pyramid!
Average 2 Minutes per Slide?
Understandability
Novelty
Cleverness
Work. Amt.
Attention
• Balance between understandability and other
conflicting metrics
• Assumption, more time improves
understandability
• Boundary condition: A single slide makes talk
more understandable?
– Sometimes more slides, graphics, animation clarify
point that otherwise is given lengthier explanation
– Amount of information is what happens
Example: Good Writing Style?
• Must describe object rather than point at it
• This does not define if the model and editor, ….
• The design pattern does not define if the model and
editor, …
• “This” should “always” be followed by a noun
– This approach, This idea, This example
• Sometimes noun is needed to formalize what one is
talking about
– Thus, the editor and model are separate components.
– This design pattern does not vs. This does not
• Sometimes noun is redundant and leads to repetition
– An alternative is to grant the access automatically under
the optimistic assumption that this will cause no harm.
– An alternative is to grant the access automatically under
the optimistic assumption that this grant will cause no
harm.
Rest are unused slides
Inductive?
You flirted with my
boyfriend
You were rude to my
mother
You kicked my cat
I HATE YOU!
Deductive?
You flirted with my
boyfriend
You were rude to my
mother
You kicked my cat
I HATE YOU!
Alternative Deductive vs. Inductive
Abstraction
• Inductive
– Not clear why information is
being given
– Too much mystery!
• Investigation described
without identifying crime
Supporting Arguments
• Deductive
– Point is clear
Abstraction
Supporting Arguments
Information Pyramid
Abstraction
Abstraction
• Give information at
different levels of
abstraction
• News paper is
• News articles can be cut at
any point from the bottom
– So can talk!
Abstraction
Question Time
• Listen to the question!
• Don’t panic
• Be honest
– “Naked presentation”
– Avoid negative comments?
• Without mentioning the
positive ones
• Self-deprecating comments
don’t work in the US
– On balance work must be
defensible!
• Be polite
– to “stupid” questions
Delivery
Understandability
Novelty
Cleverness
Work. Amt.
• Make eye contact
– Don’t look at just one person or a subset
– Look for questions and light bulbs flashing
– Though might focus on one’s asking questions
• Don’t hide slides
• Move around
– For e.g. towards person asking question
Attention
Humoyur
Understandability
• hh
Novelty
Cleverness
Work. Amt.
•
•
•
•
Attention
Relaxes people
Often insightful
In-context
Contribution obvious
only in retrospective
• Supports mystery story
Put in examples
Main Outline?
Understandability
• hh
Novelty
Cleverness
Work. Amt.
Attention
• Most talks have similar
outline
• First few slides decide if
people will pay
attention
• Don’t say something
that cannot be derived
from previous talk
– Except title
Put in examples
Principle of Good Flow
Understandability
•
•
•
•
•
•
Novelty
Problem
Related Work
Approach
Evaluation
Conclusion
Future Work
Cleverness
Work. Amt.
Attention
• Only say things that can be
derived from talk so far
– Except title
– Main outline violates this in a
minor way
Put in examples
News Example (Philip Yaffe)
Britain yesterday has once again called for the
United Nations to mount a peacekeeping
operation in the violence-torn Darfur region of
Sudan in response to increasing complaints
from aid agencies on site that international
efforts to help Darfur's desperate, displaced
population are woefully inadequate.
1.
2.
3.
4.
5.
6.
7.
The British Government is
concerned about the situation in
Darfur.
Darfur is a violence-torn region of
Sudan.
Britain believes a peacekeeping
force is urgently needed.
It is pressing the United Nations to
supply this peacekeeping force.
This is not the first time that it has
urged the UN to supply
peacekeeping force.
The population of Darfur has been
displaced.
Aid agencies in Darfur say that
international assistance to these
distressed people is inadequate.
News Example (Philip Yaffe)
At the same time, Her Majesty's Government is
joining with other European Union countries to
threaten sanctions against Sudan unless its
government energetically moves to end the
"ethnic cleansing" against black villagers in
Darfur by the mainly Arab Janjawid militias. UN
officials report that the conflict has already
claimed from 30,000 - 50,000 lives and about
1.2 million people have been displaced, with
about 200,000 taking refuge in neighbouring
Chad.
1.
2.
3.
4.
5.
6.
7.
8.
The trouble in Darfur is a race war
Arab militias are attacking black
villagers.
Britain and other EU countries believe
the Sudanese Government is not
doing enough to stop the war.
They threaten sanctions against Sudan
if its government does not quickly take
action to end the attacks.
To date, between 30,000 - 50,000
people have been killed.
About 1.2 million have been displaced.
About 200,000 have fled across the
border into the neighbouring country
of Chad.
These figures come from the United
Nations, which is a reliable source.
Nutshell-Detail-Summary
Understandability
Novelty
Cleverness
Work. Amt.
Attention
• Give basic idea
– Uses divide and
conquer
• Give solution
– Algorithm and
performance
• Summarize the topic
– Used divide and
conquer and it works as
well
Illustrate with Examples?
Understandability
Novelty
Cleverness
In the small
Grammar, PPT Animations
Style, PPT Color Choices
Analogous to defining an object
In the large
Composition of prose and slide items
Analogous to design patterns
The passive voice should
not be used!
Have an abstract,
introduction, body ,
conclusions and future
work
Work. Amt.
Attention
• Can show less work
• In soft fields or talk or
conference paper
– definition/proof by
example useful
abstraction technique
Abstraction
Abstraction
Abstraction
Multiple Levels of Abstraction
•
Title
Problem /
Issues
Related Work
Most Abstract Solution
More Issues &Details
More Issues & Details
•
•
Need to balance
abstraction and suspense
Bring existing before giving
basic solution.
Bring out more issues and
(maybe approaches if any)
before giving next level of
details
Time in Different Kinds of Talks
•
Can give results without proof in conference
talk at least
–
•
Do not need to give all details
Time Available
–
–
•
C
J DL
Shortest possible time
Potential interview
Job talk
–
–
•
Consequence can be failure
Committee knows work
Job talk precedes
Conference talk
–
–
•
But do not shy from complexity
Should describe at least one component in
some depth
Defense
–
–
–
•
Don’t shy away because of complexity
Decides future!
Conference talk precedes
Class lecture
–
–
–
Cannot afford overhead
Audience asks clarification
Job talk precedes
Indicators of good flow
• Connection to previous
slide
• Connection to far away
slide
– “The model leaves
several questions open
such as the replication of
the modules. So let us
consider these
architectural issues.”
– “This idea has the
problem/advantage”
– “The story so far is …”
“One issue I have not
addressed is”
– “I will connect this to …
later” “I had mentioned
that I would find a
connection to … This
slide does so.
Outline/Road-Map?
• Framework
• Architecture
• Outline can create flow
– Users have been told the
sequence of items
• But still need flow to
connect outline items
– “We will look at the system at
multiple levels of abstraction.
First the model, then ….”
• Re-show outline on each
context switch and recreate flow
– “That is all I will say about the
model. Let us now see the
architecture that implements
it.”
Outline?
• Framework
• Architecture
• Has components not
found in all talks
Illustrate with Examples?
Understandability
Novelty
Cleverness
Work. Amt.
Attention
• In soft fields or talk or
conference paper
– definition/proof by
example useful
abstraction technique
Related Work?
Understandability
•
•
•
•
Novelty
Cleverness


Work. Amt.
Attention

Needed to prove novelty
Contribution obvious only in retrospective
Supports mystery story
Should be covered at some level of abstraction
– Sometimes Hard to explain other works in conference
– “No other existing approach solves this problem, take
my word for it.”
• Easily possible to overdo details
Example: Conversational Style
• “The design pattern does not define
if the model and editor are (not
shown in slide)
– Centralized
– Replicated.
Centralized
• So let us consider these architectural
issues. (not shown in slide)”
• Graphics and animation improves
conversational style
– Words explain graphics on screen
– Like slide show
Replicated
Example: Conversational Style?
• The design pattern does not
define if the model and editor,
which, recall, performs input
and output, are centralized or
replicated. So let us consider
these architectural issues.
• You might as well be reading
paper.
Example: Good Writing Style?
• Must describe object rather than point at it
• This does not define if the model and editor,
which, recall, performs input and output, are
centralized or replicated. So let us consider
these architectural issues.
• The design pattern does not define if the
model and editor, which, recall, performs
input and output, are centralized or
replicated. So let us consider these
architectural issues.
• “This” should “always” be followed by a
noun
– This approach
– This idea
– This example
Example: Good Writing Style?
•
•
•
•
Must describe object rather than point at it
This does not define if the model and editor, ….
The design pattern does not define if the model and editor, …
“This” should “always” be followed by a noun
– This approach, This idea, This example
•
Sometimes noun is needed to formalize what one is talking
about
– Thus, the editor and model are separate components.
– This design pattern does not vs. This does not
•
Sometimes noun is redundant and leads to repetition
– An alternative is to grant the access automatically under the
optimistic assumption that this will cause no harm.
– An alternative is to grant the access automatically under the
optimistic assumption that this grant will cause no harm.
•
Write the noun and then remove it if necessary
– This way you know what you are talking about