perm filename LEC[AM,DBL] blob
sn#500066 filedate 1980-04-17 generic text, type C, neo UTF8
COMMENT ā VALID 00003 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 Here is the text of an informal talk I gave to Simon's UNDERSTAND
C00004 00003 I started preparing for this talk by asking my colleagues in Computer
C00064 ENDMK
Cā;
Here is the text of an informal talk I gave to Simon's UNDERSTAND
seminar last yeat, and to my CS301 class (The CS Profession) this year.
I borrowed liberally from Winston's
notes and from Strunk & White.
> If the theme of "origins of lecturing heuristics" is worth preserving
Then we must follow through (toward the end I degenerated into merely
giving country advice, without analyzing its source of power).
> In any event, much more material should probably be incorporated
>> Perhaps as an "open" project (a bulletin board of sorts)
Where a large community acesses, adds to,... a growing body of rules
>>> Maybe even tied into what Ira Goldstein is doing vis a vis the
Arpanet (Semantic net of SIGART readers' profiles)
Do you have any suggestions?
Doug
I started preparing for this talk by asking my colleagues in Computer
Science what they considered to be the most important heuristic for
lecturing.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Make sure you're wearing clothes.
-- Professor 1
Make sure your fly is zipped.
-- Professor 2
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
While there was surprising agreement in their answers, this probably
reflects more on my choice of friends than on effective lecturing.
INTERRUPT! When I stand over on this side <<META>>, I'll be making a
comment about this talk itself, a meta-level remark.
I've just opened my talk in a very weak, ineffectual way. What was
wrong with the way I did it? There are really 3 mistakes:
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
THREE MISTAKES
> OPENING WITH A JOKE
> GIVING NO CARROT
> GIVING NO OUTLINE
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
Opening with a joke is usually a disaster. In the first fifteen
minutes, people are looking for pencils, they're still arriving,
they're not responsive yet. They don't understand what you're doing
in the lecture, and by the time they realize that what you started
with was supposed to be funny, most of its impact is dissipated.
Second, and more importantly, I haven't given you a statement of what
will be learned. It's important to present such a "carrot" early,
and to be excited by it.
Third, you still have no idea what this talk is going to cover.
Somehow, I need to communicate to you a concise overview of what this
talk is all about. The best way I've found to do that is to present
an OUTLINE of the talk, and to keep redisplaying it now and then
during the hour, checking off each point as it's covered.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
OUTLINE
A WEAK OPENING
A CARROT
HEURISTICS BASED UPON I.P.S. LIMITS
MORE SOPHISTICATED HEURISTICS
Based upon Acculturation
Based upon Individual Personality
Based upon Psychosociology
MORE PRIMITIVE HEURISTICS
Based upon Physics
Based upon Physiology
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
<<check off "Weak Opening">>
Let's go back down into the talk itself, and rectify mistakes 2 & 3.
<<SHIFT out of META level>> Let me start again.
Perhaps the most common opinion about effective speaking and writing
is that they depend upon innate gifts, that they cannot be taught.
We in AI and Information Processing Psychology are in a position to
attempt to de-mystify some of this. While it's very difficult to
work out operational rules for good lecturing, we hypothesize that it
can be done. This talk will present many such "heuristics", rules of
thumb. We'll see how some of them fall right out from well-known
paramenters of the human information processing system. Some of the
heuristics have even more primitive roots, from humans as biological
organisms, or even from humans viewed merely as physical objects.
Finally, some heuristics have a much more sophisticated grounding, in
theories of personality, both individual and societal.
Not only should you come away with a sense of how these heuristics
work -- how they de-mystify effective lecturing -- but I'm hoping
that at least a few of the individual heuristics will be new to you,
that you'll be able to use them in your talks. <<Check off the
"carrot" on the outline>>
<<META LEVEL:>> Notice that this organization is far from the only
one: on first blush it might seem that the most natural thing would
have been to group all the heuristics together which dealt with
choosing a topic, all the rules for engineering sentences together,
and so on, regardless of their origins. That would certainly have
been better if my only goal were to convey some individual
techniques. I chose this outline to make it easier for me to satisfy
my other goal: to show you the straightforward origins of these
techniques.
<<Back from META>> For each of these levels, then, we'll review some
facts, some more or less well-accepted results, and based on those
facts we'll propose some heuristics for lecturing.
********************* IPS level **********************************
First, consider the level which is probably most familiar to most of
you in this room: the level at which a human is viewed as an
information processing system, a symbol manipulator, a problem
solver.
We'll assume that the human can be modelled by a set of memories...
<< IPS diagram from Newell and Simon's HPS>>
...(holding facts and methods) and internal processors, organized
just as Newell and Simon describe in Human Problem Solving. As for
the detailed characteristics of these memories and channels, we'll
accept some of the more fundamental results from cognitive
psychology:
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
SOME FAMILIAR PROPERTIES
OF THE HUMAN I.P.S.
STM
Small (7 +- 2)
Retain 2 while another task is run
Amodal
Writing: 50 milliseconds/symbol
Reading: 50 milliseconds/symbol
LTM
10
Infinite (10 >> 256k)
Associative
Amodal
Writing: 10 seconds/chunk
Reading: 300 milliseconds/symbol
EM (VISUAL)
Large
Writing: 1 second/symbol
Reading: 100 milliseconds/symbol
PROCESSOR
Serial
EIP time: 50 milliseconds
Production system
Goal-directed
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
A typical spoken sentence will span about 4 seconds, and contain
10-15 words. There's not enough time to be guaranteed of
transferring even a single symbol into LTM; 2 of the locations in STM
will probably retain some information from the last sentence after
this one is ended, meaning that only 5 symbols can be stored there.
If we don't speak the language of the speaker, then each symbol is
probably the size of a single syllable. If we speak his language but
can't make any sense out the sentence, then each symbol is probably a
word. So in that case, we could absorb and mull over a sentence of
five words or less. If we're very familiar with the language and
with the topic being discussed, then of course the words are chunked
together, and a much more complex sentence can be understood. But
note that in no case can we be expected to keep more than about 5
items on our "stack". If we have too many pieces presented before
they're tied together, some of them will have slipped away.
Look at this situation from the other side, now. Suppose you have a
thought you want to communicate. In order to maximize the number of
listeners who'll understand you, speak in short sentences, and
eliminate needless words from each sentence. These are two tenets of
an excellent book: Strunk & White's ELEMENTS of STYLE.
"Because the last few words of a sentence will be most likely to be
retained (at least for several seconds), it's important to place the
emphatic words near the sentence's end."
That sentence would have been less effective had I reversed the order
of the clauses ...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Because the last few words of a sentence
will be most likely to be retained (at
least for several seconds), it's important
to place emphatic words near the
sentence's end.
Place emphatic words near the sentence's
end, because the last few words of a
sentence will be most likely to be
retained (at least for several seconds).
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
...for then you'd think I was saying something about word retention
as a function of position; wheras I was actually recommending a
certain placement policy. This idea has a global analogue: close
your talk with a dramatic sentence: end with a joke or a memorable
phrase or both (e.g., the heuristic use of heuristic beings).
This whole problem of limited STM size is relieved by providing the
audience with an appropriate EM upon which to gaze: say a vu-graph
containing the offending sentence. This is especially useful for the
case where you genuinely do have many premises to state, followed by
a conclusion that finally ties them together into a meaningful unit.
The solution is easily understood by looking at the characteristics
of foveal vision. The same justification can be made for the
following heuristic: use vivid, concrete imagery. Prefer the
specific to the general, the definite to the vague, the concrete to
the abstract. E.g.:
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
He showed satisfaction as he took
possession of his well-earned reward.
He grinned as he pocketed the coin.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
The surest way to arouse and hold the attention of the listener is by
being specific, definite, and concrete. The great writers -- Homer,
Dante, Shakespeare -- are effective largely because they deal in
particulars, they report details that matter. Their words call up
pictures in our minds. Those pictures are almost as accessable as a
vu-graph image, as a "genuine" EM.
I believe that this same virtual EM mechanism explains the
effectiveness of examples. A good example, which the speaker keeps
referring back to as he develops his argument, is a concrete anchor,
a specific image which contains easily-accessed rich structure.
Failure to have such an archetype can at times be quite frustrating
to the listener, as e.g., right now.
Returning to our list -- actually Newell and Simon's list -- of
invariant characteristics of the human information processing system,
we examine LTM. The most macroscopic property is the slow write time
(10 seconds/chunk). If a complex idea is composed of ten lesser
ones, each of which is comprised of ten primitive chunks, then we
can't expect to learn more than 3 or 4 such new ideas in a 1-hour
lecture. This has two important corollaries:
Cor. 1: Humans can learn an idea in an hour lecture. Hence, don't
hesitate to insert an idea.
Cor. 2: Humans cannot assimilate 40 ideas/hour. If you really have
that many ideas worth communicating, you should re-think the whole
situation. Perhaps they should come out in book form, or distributed
over several articles. More likely, you'll find a couple major
themes emerging, plus an array of small related items.
<<META: >> E.g., in this lecture, we have the theme of the
explicability of effective lecturing, plus we have many individual
heuristics that are worth knowing. In such a situation, the typical
solution is to take written notes, to transfer the ideas from STM to
EM. Another common solution is REPETITION. A lecturer may place his
notes and vugraphs in the library, in order that students may review
it at their own pace. A more comprehensive kind of repetition is
frequently the solution imposed on students who don't take written
notes in class and don't review the material afterwards. More
seriously, the idea of teaching through repetition is quite
efficacious. Cycle in on difficult ideas, both within a single
lecture and in a complete course.
Note-taking is not satisfactory. Consider the tremendous cost of the
STM-->EM transfer. Probably the listener can't attend fully to the
speaker when he's furiously taking notes, certainly he can't spend as
much processing power assimilating the material. After all, he's a
uniprocessor. This dilemma explains why it's so effective to pass
out a detailed outline of the talk you're giving, so that all the
"content" is there already. Listeners need jot down only an
occasional remark or two.
<<PASS OUT OUTLINE>>
Even worse than trying to cover 40 separate points in a single hour
is the attempt to convey 40 ideas in order, where you must fully
grasp the earlier ones in order to be ABLE to understand the current
one. A common case of this vicious dependency is found in
mathematics colloquia, where what is really a large tree of
propositions is embedded in a linear order and presented that way
(lemma, proof, lemma, proof,..., theorem, proof).
The problem with that is simply that not much is going to make it
into LTM; very little will be retrievable 30 minutes later. My
objection evaporates, of course, if one is presenting a written
article. Analogously, during a talk one can write important
definitions and lemmata on the board, or distribute a handout
containing them.
The human IPS processor is able to work just a bit faster than
necessary to keep up with the 3 words/second rate of human speech.
What is happening during the rest of those cycles? Assimilation of
the new material, no doubt: tying it in to already-known concepts. I
hypothesize that if the listener doesn't understand what's said, he
uses this computational power to try to make sense out of it (to find
his own examples of it, to "punctuate" it, etc.). This processing
can be anticipated and facilitated by the speaker, by preparing
audio-visual aids which exemplify, clarify, and pucutate what is
being said. It may aid STM to duplicate exactly what's being said,
but it won't aid in understanding. For that, you must SUPPLEMENT,
not merely replicate, what's being said...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Colorful slides supplement furiously
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
...as that one does.
Another reason for slides is that often verbal description is not the
right "language" to use to convey some idea. A graph or picture may
be appropriate, or a program listing. Don't underestimate human
ability for empirical induction: often, if you can't describe some
idea well, you can still convey it by a few well-chosen examples,
non-examples, and near-misses.
Another way to rationalize the use of supplemental slides is that the
human IPS is goal-directed: it tries to understand. The aural
signals emanating from the podium are a major source of knowledge,
but so can be the visual images being projected. The synergy between
cooperating knowledge sources is much in the news these days.
If we stop focussing just on low-level, local, short-range goals of
the human, then we leave the domain of applicability of data on
characteristics of the IPS.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
If we stop focussing just on low-level,
local, short-range goals of the human,
then we leave the domain of applicability
of data on characteristics of the IPS.
If we look at some higher-level, more
abstract, global, long-range goals of the
human, then we enter the realm where
theories of personality and sociology
apply.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
<<META: >> Notice how much more effective it would have been to
phrase that sentence in the positive, to indicate where we're going,
rather than where we no longer are. An even more effective way of
conveying the idea of this shift is to put back the outline of this
talk and point at it:
<<Back from META>>
<<SLIDE: Display the OUTLINE, again>>
If we worry about all the goals of the human, then we jump from this
topic down to this one <<On the OUTLINE slide, point from IPS LEVEL
to More Sophisticated LEVEL>>. After we do that, we'll regress a
bit, and discuss some of the more nitty-gritty heuristics.
*********************** Sophisticated heurs *********************
Some of these heuristics stem from our training, our educational
experiences, our culture. For instance, we've become conditioned to
a one-hour attention span of sorts, and it's dangerous to speak
longer than that.
One important class of heuristics deal with using language to
advantage.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
EFFECTIVE USE
OF LANGUAGE
RULES OF USAGE
A participial phrase at the beginning of a
sentence must refer to the grammatical subject
RULES OF COMPOSITION
Choose a suitable design and hold to it
Use the active voice
Put statements in positive form
Place negative and positive in opposition
* Use definite, specific, concrete language
* Omit needless words
Express co-ordinate ideas in parallel form
Keep related words together
Keep to one tense
* Place the emphatic words at the end
RULES OF FORM
Worry about colloquialisms, hyphens, exclamations
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
We've already discussed "omit needless words", and "place emphatic
words at the ends of sentences", and "use definite, specific,
concrete language". There are many more:
Some are rules of USAGE ("a participial phrase at the beginning of a
sentence must refer to the grammatical subject")
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Being in a dilapidated condition, I was able to
buy the house very cheap.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
Some are rules of COMPOSITION: "The active voice is usually more
direct and vigorous than the passive". Similarly, "put statements in
positive form." Placing negative and positive in opposition makes for
a stronger structure.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Not charity, but simple justice.
Not that I loved Caesar less, but that I loved Rome more.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
Express co-ordinate ideas in similar form, i.e., use parallel construction.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
A heuristic is not a rule of inference, merely a rule of thumb.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
Keep related words together...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
New York's first commercial human-sperm bank opened Friday
with semen samples from 18 men frozen in a stainless steel tank.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
... In this version of the sentence, the listener's heart would go
out to those 18 poor fellows frozen in a steel tank.
Some high-level rules of FORM exist: take care how your audience
might misinterpret various sentences, ideas, expressions. In
Chattanooga, when 2 newspapers merged, the NEWS and the FREE PRESS,
they blindly hyphenated them to get the new paper's name: the
Chatanooga News-Free Press.
This set of rules is of course more exemplary than exhaustive; as
Strunk and White put it, who can say confidently what ignites a
certain combination of words, causing them to explode in the mind?
Every speaker, by the way in which he wields language, reveals
something of his spirit, his habits, his capacities, his biasses.
This is both inevitable and enjoyable. But don't confuse spontaneity
with genius. A breezy style, well maybe an iota more breezy than
mine, say, is the mark of an egocentric. In general, if you're going
to deviate from the cultural norms, from accepted rules of grammar
and style, do it for a reason in each case: be intentionally,
purposivley provacative or ingenious. Perhaps the highest ideal at
this level is clarity. When you become hopelessly mired in a
sentence, stop and start afresh: don't try to hack your way out
against the terrible odds of syntax; usually what's wrong is that the
sentence construction has become too involved at some point. Start
over, using much shorter, simpler sentence constructions.
When you say something, make sure that you have said it. The chances
of your having said it will then be at least fair.
Avoid jargon as much as possible, even in a technical talk. We tend
to drag in a pouch of private words whose only virtue is that they
are exceptionally nimble and can escape from the garden of meaning
over the wall. Learn to spot words that at first seem frieghted with
meaning, but then soon burst in air, leaving nothing but a memory of
bright sound. Even worse than jargon buzzwords are
PROGRAM-VARIABLE-NAMES, using the actual program's print-names, no
matter whether they're the best possible names or not. Unless you're
giving a minicourse on how to run and modify a particular piece of
software, it can only help to rename variables so as to be maximally
suggestive.
Perhaps the ultimate customizing method at this "word" level is
sentence engineering: rework a sentence until it will have meaning
at several levels, so that everyone in your expected audience will
think they understand it.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
A heuristic is really just compiled hindsight.
A.M. creates new concepts and explores facets of them.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
Note that the first sentence can be understood by someone unfamiliar
with the technical computer science meaning of compilation, though of
course one gets more out of the sentence if he has and uses that
meaning. In the second sentence, one can still extract some useful
meaning, even he doesn't know that concepts and facets refer to a
particular, well-defined data structure in the A.M. system.
Always think of your audience, both locally when you engineer
sentences, and globally when you plan your talk.
It is important to realize that a lecture for CS104, introduction to
programming, is quite different from the Turing award speech. By
isolating the overall goals of the speaker and audience, we can
categorize a few situations worth distinguishing.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
Technical
Peer Audience
Job Interview
Invited Talk
Casual
How I Spent my Summer Vacation
Popular
AI for Tots
After-dinner Speech
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
In a continuing course, the instructor must bear in mind his general
course goals, and each lecture must fit into this high-level plan.
In a technical seminar, it would be disastrous to under- or
over-estimate the level of expertise of the audience. At a talk
you're giving to get a job interview, the goals are obviously quite
different than if you've been asked to make a few remarks upon
accepting some award. When you compose and edit your talk, you must
filter it past the goals: does it satisfy your goals? the
audience's?
MY purpose...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
TO MOTIVATE
Show Results
Entertain
jokes
question the audience
intentionally trigger questions
home demonstration
visual effects, props, gimmicks
multi-media, multi-screen
gestures
eccentricity
eye contact
approach audience
approach screen
professional graphics
emergency time-sink
crowded rooms
don't get deflected
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
... in delivering talks -- at least, talks about my research -- is to
both convey knowledge, and to get the audience interested.
Recognizing that little knowledge can be permanently transferred in
an hour, I concentrate on motivating. If the listener gets
interested, he or she will talk to me in more detail, go and read
some articles on the project, and thus -- in TENS of hours -- acquire
a grasp of what I'm working on.
How do you motivate an audience? Part of it is to show them some
interesting results, to arouse their curiosity in HOW those results
were acheived, to convince them that there is some power here if
they're willing to study it. Another part of motivating is
entertainment, both for its own sake and to assure the listener that
should he choose to pursue this, he'll find further material which is
comprehensible and enjoyable.
Be careful about jokes and puns: if they fail it's a real disaster
(unless you are skilled at appealing to the audience for sympathy).
But a few good jokes can brighten a talk tremendously. Knuth's
criterion for a good joke is that it be TECHNICAL, that it require
some detailed understanding of the current material, in order to get
the joke...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
TECHNICAL JOKES
On maps, thick red lines constrain furiously.
The heuristic use of heuristic beings.
The "Rock Test" parody of Turing's test.
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
...I can think of a couple reasons for this: (1) When we get the
joke, we reward our own cleverness -- and publicize it -- by
laughing. (2) As Arthur Koestler says, the discovery of the jarring
of two separate matrices of thought is what creates the appreciation
of humor, and in this case we have two simultaneous such jarrings,
one for the joke itself, and one because the joke is highly relevant
to the current discussion.
A technique which, like jokes, can work well or can backfire, is
asking questions of the audience. It helps convert a passive
situation into an active participation.
An even more subtle technique is to arrange your discourse, and even
your slides, in such a way that a particular question is just begging
to be asked. If that question is on several people's minds, and
especially if someone voices it, then it will be extremely effective
when you answer it, and in fact when that answer flows into the talk.
An excellent way to motivate is to propose some little self-contained
experiment or demonstration they can try later, on their own, to
confirm a point of the lecture.
Most talks are delivered in such a routine way that any variation at
all is likely to sustain interest. Use visual effects, especially
props and complete demonstrations.
Get a few emergency tricks, in case you feel an audience slipping
away. Try dropping a rock, or asking people to keep their eye on this
point <<at this moment, place a paper cut-out of a Xerox image of my
hand (pointing) on the vu-graph, creating the illusion that I'm
pointing; then walk away and continue talking non-chalently>>.
Try appealing to several senses (almost any recording in the
background is a novelty), try composing some of the above or
compounding them (3 slide projectors); (writing [esp: with colored
chalk] on vu-graph image projected onto the blackboard instead of
onto the screen).
Cultivate gestures. Point at the board a lot. This is effective
even if what you point at is unrelated to what you're saying.
Have an eccentricity. Chew tobacco, wear three piece suits <<I'm
currently in a three piece suit as I say this>>, or point at the
board a lot.
Look people in the eye often. This establishes that you are talking
with them, rather than presenting material as if to a videotape
machine.
Try to break down the natural barrier that exists between speaker and
listener. Walk toward the audience and away from them, as well as
left to right across the room. <<Act this out, of course>>
Get physically close to any projected images. <<Act this out>> Aside
from the entertainment aspect, this has a god IPS justification: it
doesn't force people to divide their attention between you and the
screen.
Similarly, don't distract people by showing them a half-covered
slide. They'll divide their attention, spending some time trying to
guess what's underneath. <<Demonstrate this>> Instead, I find it's
much better to just use overlays: that way, the audience doesn't know
that anything's coming until it's there already.<<Overlay: "See!">>
Neat drawings and slides, particularly in color, create the illusion
that the lecturer is at least twice as well organized as he really
is. At IJCAI, several of my slides were not quite legible, except
for their headings. Yet, because they -- and the rest of my talk --
seemed so well organized, many people came up to me and congratulated
me for calculating so effectively just what material on each of my
slides was supposed to be legible, and what was intentionally
illegible.
On the same point, carry a pipe or a half-empty coffee cup. That
way, if you get stuck, stop and think while puffing or sipping.
<<Have the coffee cup ready for this point, of course>>
Many of the entertainment heuristics had to do with mass psychology.
There are some more along those lines that we should all know:
Avoid sparsely populated lecture halls. Empty seats suggest, whether
true or not, that attendance is much lower than expected. Just a few
people forced to stand, <<as in this case -- hopefully>> on the other
hand, suggests the opposite. Steeply sloping halls are also bad, like
ScH 7500. They give the effect of standing at the bottom of the
grand canyon, lecturing to the stones.
A group is emotive in proportion to its size. A huge audience will
have extreme positive and negative reactions. There is a noticable
difference even between 50 and 150 people.
If you get a deflectionary question, deflect it. Don't allow
obstructionist questioners to drag you off your plan. Talk with them
after the lecture, if they wish.
*********************** Primitive Heuristics ********************
<<Display the slide with the OUTLINE AGAIN>> We're now going to move
on to some heuristics which are based on very primitive views of man.
First, consider Man viewed as a physical entity...
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
OTHER VIEWS OF MAN
AS A PHYSICAL ENTITY
Mass
Opacity
Unidirectional vocalization
Unidirectional motion
....
AS A BIOLOGICAL ENTITY
Optical resolving power
Aural receptive strength
Need for food
Limited rate for talking
Need for sleep
ESP absent (or undeveloped)
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
An important corolloary to the second point listed (opacity) is that
Man is not transparent to vu-graph light. If you stand in front of
the lens, the audience will no longer see the image. I never
understood how speakers could do such a thing; I have a very simple
heuristic: I know I'm blocking the projector iff I notice a blinding
white light in my eyes <<Demonstrate>>. I just found out yesterday
what the problem was, when Bentley told me HIS analogous heuristic:
he knows he's been blocking the projector for five straight minutes
iff his chest begins to feel warm. So this turned out to be just a
5'6" heuristic after all.
Seriously, we all forget at times that what the audience sees is the
screen, NOT what we see illuminated when we look down on the
vu-graph. A specific danger is that the scales of graphs will just
barely miss being projected. <<Demonstrate this>> As for
unidirectionality of projected speech, well, make the obvious
correction when you turn away from the audience.
If we view man as a biological organism, then we must cater to his
physiological limitations, just as before we viewed him as an IPS and
catered to his information processing limitations. In this case, we
must worry about the resolving power of his eyes <<tiny font>>, the
<<quietly>> sensitivity of his ears, and so forth. Our limited
speaking rate has imposed upon us a READING rate limit (Evelyn Wood
would say, artificially imposed). <<too much print to read>>, so
don't put too much on any one slide. Humans need to sleep
periodically. Not only does this put an upper limit on how long any
speaker could possibly take (about 1 week straight), it gives the
speaker some danger signs to watch for (napping, yawning). Finally,
since humans have low ESP abilities (at least, not developed much),
it is generally a good practice to define your terms before using
them. Also, due to low ESP abilities, people won't know what you're
omitting from a talk. If you run overtime, be merciless in cutting
out material. Get right to your conclusions. I'm sick and tired of
having a thoughtless lecturer run 40 minutes overtime, when most of
his audience is still there only out of courtesy.
<<META: >> In this last section, I've switched my attitude from
friendly to antagonistic. Notice, by introspection, now, how it's
alienated you a bit, if you picked up on that tone of hostility and
condescension. This illustrates a very important heuristic, valuable
enough to risk having you leave here angry at me: never make your
audience angry at you; go out of your way to show your good faith.
Even if you know a lot about the subject you're speaking on, never
forget that they know more about a host of other topics. So don't be
condescending; treat your audience with respect. Adopt the attitude
that you will do ANYTHING to help them understand, that you are not
doing them a favor, that they are worth your time. That attitude
will project, and if you stumble the audience will support you.
Just a few words about closing. One philosophy says "tell them what
you're going to tell them, tell them, tell them what you told them".
One unoffensive way to do the latter is to put the original outline
back up, so they can see what's been covered. <<Do this>> Another
technique I use is to put up a new slide containing the main points
or themes, or containing the "open problems" left to worry about.
This is a good slide to leave up during the ensuing question and
answer period.
[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
GOALS OF THIS TALK
TO CONVEY INFORMATION
Heuristics exist for effective lecturing
Their origins are straightforward
TO MOTIVATE
Study and master the heuristics
Investigate rule-guided communication
BIBLIOGRAPHY (v. small font)
Bentley, Jon. personal communication
Feigenbaum, Edward. personal communication
Shamos, Michael. personal communication
Strunk, W., and E. B. White, The Elements of Style, MacMillan, 1972.
Winston, Patrick. personal communication
]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]
As to final sentences: it often is appropriate to close with a joke,
or with a memorable phrase ("the science of science"). The important
thing is to know when to close. It's horrifying to listen to someone
waste five minutes out of some compulsion to fill a time slot. Stop
when finished <<stop abruptly>>.