Trees, maps, and theorems: sample pages

gaywaryΗλεκτρονική - Συσκευές

8 Οκτ 2013 (πριν από 3 χρόνια και 10 μήνες)

137 εμφανίσεις

Trees,maps,and theorems
Effective communication for rational minds
Jean-luc Doumont
© 2009 by Jean-luc Doumont
All rights reserved
Published in 2009 by Principiae
in Belgium(www.principiae.be)
First printing,January 2009
isbn 978 90 813677 07
d/2009/11.719/1
About this book
I
nquest of solidwrittensupport for the participants
of my training sessions,I searched libraries,bookstores,
and mail-order catalogs,but to no avail:I did not find
a reference that quite matched the approach I had developed.
Encouraged by the feedback on my lectures and publications,
I thus set to create my own book on effective communication,
for my usual audiences of engineers,scientists,andmanagers.
The outcome of this endeavor is the book youare nowreading.
This book is about first principles.It is about strategy and,
especially,about structure.To borrow Hemingway’s words,
it is about architecture,not interior decoration.It is about
constructing communication deliberately and methodically.
It is about reaching a given purpose with a given audience,
in virtually any professional situation—and in any language.
This book is for professionals who want to master the basics,
that is,to understand themclearly and apply themcarefully
when communicating on the job.It is for those who believe
that effective communication skills are an invaluable,lifelong
personal asset and who want to keep strengthening this asset.
As such,it benefits students,too,notably graduate students.
This book,however,is no self-study course—no book can be.
Sharpeningone’s skills requires practice onone’s ownmaterial.
Moreover,it requires feedback,for practice without feedback
is of little use.Global feedback may come out of the situation
(Did I get my message across?).Acareful analysis,in contrast,
requires an instructor or mentor—a human being,not a book.
This book has been described both as a minimalist user guide,
with its concise instructions,carefully selected applications,
and answers to frequently asked questions,and,interestingly,
as a children’s book,with its precise yet straightforward tone,
its exposition of one topic per double page (most of the time),
and its illustrations.These two descriptions are fine with me.
Onwhatdoyoubaseyourrecommendations?
Theguidelinesinthisbookarebasedmostly
oncommonsenseandexperience.Theyhave
beenputtothetest,notonlyinmyownpractice,
butalsobythousandsofengineers,scientists,
managers,andotherprofessionalsworldwide
whotookpartinsomeofmytrainingsessions.
Ihopetheguidelinescanbeasusefultoyou
astheyapparentlyaretotheseprofessionals.
Moreover,myapproachisnodoubtinfluenced
bymyeducationasanengineerandscientist,
and—inwaysdifficulttotraceortoquantify—
byallIhavereadorheardoncommunication.
Doyourelyonempiricalresearchatall?
Well-conductedresearchinanyscholarlyfield
isnormallythought-provokingattheveryleast,
soresearchfindingsshouldnotbedisregarded.
Still,empiricalresearchaboutcommunication
suffersfromverymanyconfoundingfactors
andisthushardtogeneralizetowardpractice.
Inmyexperience,fartoomanypeopleapply
poorlyunderstoodresearchoutcomesblindly,
sometimestotheextentofgeneratingmyths.
Iwouldratherthattheythoughtforthemselves.
Whysuchafocusoncounterexamples?
Remarkably,thereisnothingquiteremarkable
aboutinstancesofeffectivecommunication:
theydrawone’sattentiontotheideasexpressed,
nottothemselves,sotheyarehardtolearnfrom
byimitationwithoutthecontrastingviewpoint
providedbyacounterexample.Also,learning
topinpointshortcomingsinone’sownpractice
isanecessarysteptowardimprovingonthem.
HowtousethisbookThisbookwasdesignedtoproposealogicalflow
forthediscussionwhileenablingselectivereading
ofindividualparts,chapters,orsections.Feelfree,
therefore,toreadthecompletediscussionlinearly
ortojumpaheadtothethemesofinteresttoyou.
Topicsarediscussedinonedoublepageeachtime
(orinasmallintegernumberofthem),tofacilitate
theirdirectaccessorout-of-sequenceprocessing.
Thepages,too,areformattedforselectivereading.
Therightpageisreservedforthemaindiscussion,
withillustrations,limitedexamples,orcomments
placedleftofthetext.Inrelationtothisdiscussion,
theleftpageanswersfrequentlyaskedquestions
collectedattheoccasionoflecturesandworkshops,
setonagraybackground.Intheremainingspace,
itliststypicalshortcomings,offerspracticaladvice
onspecificsubtopics,orbroadensthediscussion.
Frequentlyasked
questions
Commonshortcomings,
practicaladvice,etc.
Illustrations,
comments
Maindiscussion
Thisbookisorganizedinfiveparts:first,fundamentals,then
writtendocuments,oralpresentations,andgraphicaldisplays,
andfinallyapplicationtofivemorespecifictypesofdocument.
Itendswithnotesandreferences,aswellasanindexoftopics.
Partone,Fundamentals,introducestheideasthatunderpin
thefoursubsequentparts.Probablythemostarduouspart
ofthebookasitlackstheexamplesthatappearfurtheron,
itcanbeskimmedorperhapsskippedatfirstbythereader
eagertostartworkondocuments,presentations,ordisplays.
Still,itanswersmanyawhyaboutfurtherrecommendations
and,byofferingaminimalsetofuniversalprinciples,itequips
readersformostchallengesofprofessionalcommunication.
Parttwo,Effectivewrittendocuments,offersamethodology
infivestepstoproceedfromscratchtoacompletedocument,
namelyplanning,designing,drafting,formatting,andrevising
thedocumenttobecreated.Itdetailseachofthesefivesteps.
Partthree,Effectiveoralpresentations,proposesasimilar
yetdistinctapproachinfivesteps:planningthepresentation,
designingit,creatingtheslides,deliveringthepresentation,
andansweringquestions.Thoughmeanttostandonitsown,
itdoesnotrepeatuselesslywhathasalreadybeendiscussed
indetailaboutwrittendocuments,inparticularplanning.
Partfour,Effectivegraphicaldisplays,firstclassifiespictures
asawaytohelpreaderschoosetherightrepresentation,then
discusseshowtoplan,design,andconstructoptimalgraphs,
andfinallyhowtodraftacaptionthatgetsthemessageacross.
Partfive,Applications,illustrateshowthegeneralguidelines
inthepreviouspartsapplytofivecommontypesofdocuments.
Specifically,itexaminessetsofinstructions,electronicmail,
Websites,meetingreports,andscientificposters,eachtime
particularizingearlierrecommendationsoraddingnewones.
Contents
Fundamentals
The name of the game 3
The three laws of communication 5
A thousand words,a thousand pictures 13
Chains and magical numbers 17
Trees,maps,and theorems 23
Effective written documents
Planning the document 33
Designing the document 43
Drafting the document 59
Formatting the document 71
Revising the document 77
Effective oral presentations
Planning the presentation 87
Designing the presentation 89
Creating the slides 97
Delivering the presentation 105
Answering questions 117
Effective graphical displays
Understanding pictures 123
Planning the graph 131
Designing the graph 133
Constructing the graph 145
Drafting the caption 149
Applications
Effective instructions 153
Effective electronic mail 157
Effective Web sites 161
Effective meeting reports 165
Effective scientific posters 169
Needparagraphsalwaysbeparallelorserial?
Paragraphsneednotalwaysbeentirelyparallel
orentirelyserial.Theymayuseacombination
ofthetwostructures,orbe“pseudo-parallel”,
(liningupcomparableyetnotidenticalsubjects).
Tobereadable,however,theyshouldnotmiss
opportunitiesforaparallelorserialstructure,
suchasintroducingaswitchinsubject(a→c)
thatdoesnotreflectaswitchintopic(yielding
a–bc–a,insteadoftheparallellinka–ba–c)
orpositioninganewitem(c)beforeanitem(b)
mentionedintheprevioussentence(yielding
a–bc–b,ratherthantheseriallinka–bb–c).
Istheparallelstructurenotboringtoread?
Parallelismmayseemtoencouragerepetition.
Notso,however:unpleasantrepetitionsmust
ofcourseberemovedlesttheybecomenoise,
butnotbyuncalled-forvariationsinstructure.
Whenattemptingto“parallelize”aparagraph,
youcanremoveresultingrepetitionsbyusing
pronounsandbycombiningrelatedsentences,
notunlikerewriting3ax+5ayasa(3x+5y).
HowshouldIpunctuateadisplayedlist?
Therulesforpunctuatingdisplayedlistsvary
frombooktobook(andlanguagetolanguage).
Whicheveryoudecidetoapply,beconsistent.
Forwrittendocuments,consistencysuggests
usinginliststhegeneralrulesofpunctuation
andcapitalization:thus,capitalsandperiods
forfullsentences,andcommasorsemicolons
toseparatephrasesorclauseswithinasentence.
Fororalpresentations,thedesiretobevisual
maysuggestdroppingthepunctuationmarks
inlists(andperhapsinsomeothertextitems).
DraftingeffectivelistsCommonnotonlyinwrittendocumentsbutalso
onoralpresentationslides,liststoooftenexhibit
shortcomingsthatrenderthemplainlyineffective.
Listsarefordisplayingcomparableitemsinaway
thatencouragestheircomparisonormemorization,
notformakingaloosesetofitemslookorganized.
Whethertheyaredisplayed(withorwithoutbullets)
ortypesetaspartofasolidparagraph,listsshould
comprisefewitems(inotherwords,fiveorfewer),
toallowtheirnonsequential,visualprocessing;
introducetheitemsbyaclause(orpartofone),
toletthereadersknowwhatthelistisabout;
phraseallitemsinagrammaticallysimilarway,
toreflectintheformtheparallelismofcontent.
Themannerinwhichtheitemsarephrasedshould
obviouslybeagrammaticallycorrectcontinuation
oftheintroductorycomponent.Theuseofbullets
torevealitemsdoesnotaltertherulesofgrammar.

Toprepareameeting,defineitspurpose
Youmustalsoprepareanagenda
Everyoneshouldreceivethisagenda
Doeseveryoneknowwhotheothersare?
Thechairpersonshouldnotbesecretary
Groundrulesmaybeappropriate,too
Alwaysreviewthepurposeandagenda
Whenpreparingameeting,
definethepurposeandagenda,
sendtheagendatoallparticipants.
Asyoustartthemeeting,
welcomeandintroduceparticipants,
clarifytheroles(chair,secretary,etc.),
setupgroundrulesifappropriate,
reviewthepurposeandtheagenda.
62
Aboveallelse,sentenceswithinaparagraphareconnected
bycontent:oneelementinasentence,normallyitssubject,
pointstoanelementintheprevioussentence.Areference
totheprevioussubjectisbestdonewithapersonalpronoun
(it,they,we,etc.),whereasareferencetothepreviousobject
orotheritemappearingattheendoftheprevioussentence
isbestindicatedbyademonstrativeadjective(this,such,etc.).
Byanalogywithelementaryelectricalcircuits,wemightcall
thefirstcaseaparallellinkandthesecondcaseaseriallink.
Parallelorseriallinkscanberepeatedforseveralsentences:
aparallelstructurelinesupsentenceswiththesamesubject,

Recentyearshaveseenanincreasedpopularity
ofcodesbasedontheDiaboloalgorithm.Speed
isamainadvantageofthesecodes,comparedto
thetraditionalDemonones.Also,onecanimple-
mentthemreasonablyeasily,anditispossible
toextendthemsotheycanhandlehybridtrans-
forms.Ontheotherhand,theyrequireabout
45%morememory,butthisislesscriticalwith
today’sarchitectures.Typicalapplicationsare…
Inrecentyears,
codesbasedontheDiaboloalgorithm
havebecomeincreasinglypopular.Comparedtothe
traditionalDemoncodes,
they
areabouttwiceasfast,
arereasonablyeasytoimplement,andcanbeextended
tohandlehybridtransforms.Asadrawback,
they
require
about45%morememory,alesscriticallimitationwith
today’sarchitectures.Typically,
they
areappliedto…
whereasaserialstructurechainssentencesbyusingwhatis
introducedinonesentenceasthesubjectofthenextsentence.

AllcurrentimplementationsoftheDiaboloalgorithm
arebasedontheso-calledAngeltransform.F.Angel
firstdescribedthistransformin[2].Theideais
toseparatethedataintohighandlowvaluesbefore
proceedingwithgeneration.Theimplementation
thenstoresthehighandlowvaluesseparately…
AllcurrentimplementationsoftheDiaboloalgorithmare
basedon
theso-calledAngeltransform
.
Thistransform
,
firstdescribedbyF.Angel[2],separatesthedatainto
highandlowvalues
beforeproceedingwithgeneration.
Thehighandlowvalues
arethenstoredseparately…
Bothstructurescanbeusedtoconstructentireparagraphs.
Theparallelstructure,usingtheparagraph’stopicassubject
ofallitssentences,isthemorereadilyapplicableofthetwo.
Theserialstructureiswellsuitedtointroductoryparagraphs,
organizedfromgeneraltoparticular,andtosubstructures,
forexamplewithinamorecomplexparallel–serialparagraph.
parallellink
Thenozzle
includesascatterer.
It
iseasytomount.
seriallink
Thenozzleincludes
ascatterer
.
Thisscatterer
is
easytomount.
Effectivewrittendocuments63
Shouldeveryslideconveyamessage?
Ideally,everyslideindeedconveysamessage,
especiallyforashort,intensivepresentation.
Ifyoudecidetoshowaslidetoyouraudience,
surelyyouaretryingtotellsomethingwithit.
ShouldInotincludebibliographicalreferences
onslides,toshowwherethedatacomefrom?
Fullbibliographicalreferencesareoftennoise.
Doyouexpectaudiencememberstoreadthem?
Ifyes,thendonotexpectthemtolistentoyou.
Ifno,thenwhydoyouincludethereferences?
Referencesareimportantinformationindeed
butarebestplacedinacompaniondocument.
Togivecredittosomeoneforthedatayoushow,
simplyincludetheperson’snametogetherwith
theyearofpublicationinHarvardcitationstyle,
asin“Doumont,2005”—notthefullreference.
HowcanIdisplaycomplexinformation,
suchasequationsorintricatediagrams?
Beforeyoulookforsolutions,questiontheneed
fordisplayingcomplexinformationonslides:
typically,suchinformationcanbepresented
moreclearlyinacompaniondocumentinstead.
Ifyoumustconveytheinformationonaslide
afterall,provideaglobalviewbeforemoving
tothedetails.Forexample,identifyandlabel
themain“blocks”ofyourequationordiagram.
∇×B=μ0J+μ0
0
∂E
∂t
current
term
field
term
MostslidesjustincludetoomuchtextAudiencemembersknowit:mostslidesoutthere
includetoomuchinformation(astextorotherwise)
forthemtoevenstartprocessingitwhilelistening
tothespeaker.Strangely,whenthesesamepeople
becomespeakers,theytoocreatecrowdedslides.
Whyso?Threeexplanationsreadilycometomind.
First,manyspeakerscreateslidesforthemselves,
asapersonalmemoryaid,notfortheiraudience.
Theseslidesareoftencryptic(notself-explanatory)
andtext-heavy.Suchmaterialmayaidthespeaker
inpreparingorevenindeliveringthepresentation,
butitshouldsimplynotbeshowntotheaudience.
Second,adrivetowardefficiencypushesspeakers
tothinktheirslidesmustdoubleaswrittenreport.
Alas,slidesdesignedwithsuchapurposeinmind
tendtoincludetoomuchtobeeffectiveasslides,
yetprobablytoolittletomakeaconvincingreport.
Inmostsituations,theyfallshortofbothobjectives.
Third,speakerswhocreatetheirslidesinahurry
oftenusematerialcopiedfromwrittendocuments
(paragraphs,spreadsheets,etc.)withoutadaptingit.
Theyknowtheresultislessthanperfectbutseeit
as“betterthannothing”.Inasmuchastheydistract,
suchslidesareinrealityworsethannoslideatall.
Thethreereasonsabovemustbecomplemented
bytwofrequentconfusions.First,somespeakers
concludethat,tostandontheirown,slidesmust
includeprettymucheverythingtheysay.Notso:
whatappearsontheslidemustbeself-explanatory,
butnoteverythingsaidneedstoappearonaslide.
Second,andforfearofstatingfactsoutofcontext,
speakerssometimesincludeextradataontheslide;
forexample,theymentionthreenumbersbutshow
adetailedtable,toallowcomparisonsifdesired.
Theaudiencecannot(andshouldnot)spendtime
analyzingalargetablewhilelisteningtoaspeaker,
however;suchalargetablebelongsinahandout.
98
DesigningtheslidesA
fteryouhavestructuredthebodyofthepresentation
inmainpointsandsubpoints,geteachsubpointacross
withoneslide—andoneonly,atleastinfirstapproximation.
Statethesubpoint’smessageontheslideasashortsentence
(inthetitlearea):expressthesowhat,notmerelythewhat.
Illustrateyourmessageasvisuallyaspossible,limitingtext.
Questionthepertinenceofanyitemyouintendtoinclude;
ifyouarenotgoingtomentionit,donotputitonyourslide.
Adraftbyhandmaysaveyoumanyiterationsatthecomputer
andhelpyouidentifywhatyouneedinsteadoftryingtouse
whatyouhave.Firststateeachofyourmessagesasasentence
andsketchanillustrationforit,andlaterconstructallslides.
Toverifywhetheryourslidesstandontheirown,showthem
tosomeonerepresentativeofyouraudiencewithoutproviding
yourspokentext.Thissomeoneshouldbeabletofigureout
whateachslidedisplaysandwhyyouareshowingtheslide
aspartofyourpresentation.Printingtheslidesinsmallsize
(typicallysixonapage)allowsyoutotesttheirlegibility,too:
whateverishardforyoutomakeoutonsuchaprintedpage
willlikelybehardforanaudiencetomakeoutonthescreen.
Theslideaboveispoorlydesigned.First,itconveys
nomessage:thetitleexpresseswhatthedataare,
notwhatthedatamean(thewhat,notthesowhat).
Second,itisnotasvisualasitseems:thelegend
ofthepiechartisanarbitrarydictionaryofcolors,
hardtoprocesswhileonelistenstoaspokentext.
Finally,itisnoisy:thepiechart’sthirddimension
distortstheperceptionofthedata,thereference
atthebottomwilleitherdistractornotberead,etc.
Mainmessage
Mainpoints
Subpoints
Conveyeachsubpointwithaslide
DISTRIBUTIONOFTHENUMBEROFSLEEPHOURSFOR
ADULTS
L’hebdoauféminin106,p.3,November12,1991
28%
41%
22%
9%
5HOURSORLESS
6HOURS
7HOURS
8HOURSORMORE
Only28%ofadultssleep
therecommended8hours
8hours
ormore
5hoursorless
6hours
7hours
28%
41%
22%
9%
Effectiveoralpresentations99
Canadisplaybecometoocomplex?
Justlikeanyothercommunicationcomponent,
adisplaycanbetoocomplexforitsaudience
orinviewoftheconstraintsofspaceandtime.
Still,whatmakesdisplaystoocomplicatedis
mostoftenasuboptimalspatialarrangement
ofitspanels.Tokeeptheperceptionglobal,
alignthepanelsvisuallyinmeaningfulgroups
ofuptofive—bothhorizontallyandvertically.
Whatistheoptimalsizeforagraph?
Findingthemostappropriatesizeforagraph
isnotautomatic:onemusttakeintoaccount
notonlythedatadisplayedbutalsothepage
orslideonwhichthegraphmustbeinserted,
toachieveaneffectiveandharmoniouswhole.
Graphsinreportsandonslidesareoftendrawn
unnecessarilylargeforthedatasettheydisplay,
wastingspacethatcouldbeputtobetteruse,
suchashavingtextnexttothegraphinareport
orlargermarginsaroundthegraphonaslide.
Asaroughguide,questionthesizeofgraphs
thattakeupmorespacethanyouwouldneed
foratablelistingthesamedataasinthegraph.
Attheotherendofthespectrum,smallgraphs
areseldomtoosmallforthedatatheydisplay
butmaynolongeraccommodatelegiblelabels
ordistinguishabledatamarkers,inparticular
ondevicesthatsufferfromlimitedresolution,
suchasprojectedcomputerscreens.Limiting
textlabels,numbers,anddatamarkerscanhelp.
Beyondthesizeofthegraphinitself,optimize
suchparametersasfontsizeandlinethickness,
especiallyifyouscalethedisplayupordown.
Manyslideshavehugegraphs…withatinyfont.
ConstructingcomplexdisplaysComplexdatasets(thoseincludingmanyvariables)
mayrequirecomplexdisplays,withenoughpanels
tosearchforcorrelationsorcompareevolutions
amongcontinuousvariables,ortodisplaysubsets
resultingfromthepresenceofdiscretevariables.
Thesepanelsmustthenbemeaningfullyorganized
onthepage,toshowthestructureofthedataset:
panelswithacommonhorizontalorverticalscale
arethusbestplacedunderorbesideoneanother,
respectively(withthescaleappearingonlyonce),
panelsshowingsubsetsmustusethesamescales,
etc.Principlesofproximity,similarity,prominence,
andsequenceapplyhereastheydoforpagelayout:
thedisplaymustindeedprovideinsightasawhole.
Placebo3mgb.i.d.6mgb.i.d.
Males
Females
Incontrast,datasetsincludingbutfewvariables
seldomrequirecomplex—orevenlarge—displays,
evenwhentheyincludemanyvaluespervariable.
Effectivegraphicaldisplayscanconveymanydata
inasmallspaceandcanoftenbereducedinsize
withoutlossofclarityaslongaslabelsarelegible.
Smallgraphsareparticularlyappropriateaspanels
incomplexdisplays,wheretheirsizeallowsricher,
clearerspatialarrangements.Suchsmallmultiples
areusuallyclearestwithasfewlabelsaspossible.
Theycanoftendowithoutthescales,forexample.
Thesamplehistogramsatright
takeuplessthan0.5cm
2
each
(andcouldbedrawnsmallerstill),
yettheystillallowcomparisons.
142
Comparinggroupsofdata
D
iscretevariables,dividingthedatasetintosubsets,
canberepresentedineitheroftwofundamentalways:
thesubsetscanbeshowninasinglepanelanddistinguished
byavisualdifferencesuchasmarkershapeorlinethickness,
ortheycanbeshowninasmanyseparate,juxtaposedpanels.
Asinglepanelallowsamoreaccuratecomparisonofsubsets
butmaynotprovideasatisfactoryviewofindividualsubsets
whendatapointsarenumerousandsubsetslargelyoverlap.
Whenthedatasetinvolvesmorethanonediscretevariable,
theresultingdisplayscanusemultipledevices,forexample
differentmarkershapes,eachinsolidandhollowversions,
ormultiplepanelsorganizedbothhorizontallyandvertically.
Theycanofcourseuseacombinationofthetwoapproaches.
Multiplepanelsrepresentingsubsetsofthesamevariables
mustusethesamescalestoofferameaningfulcomparison.
Bycontrast,multiplepanelsrepresentingdifferentvariables,
asinamatrixplot,mayuseadifferentscaleforeachvariable,
forexampleinanefforttoresolvecloselygroupeddatabetter.
t=0
t=1
t=2
Singlepanel,
multiplelines
t=0
t=1
t=2
Multiplepanels
Fasterdiffusion
Fasterdrift
747576
MalesLifeexpectancy
France
Germany
Netherlands
UK
Belgium
80818283
FemalesPopulation[millions]
Germany
82.2
UK
59.6
France
58.7
Netherlands
15.9
Belgium
10.2
Density[inhab./km2
]
230
244
108
467
335
Effectivegraphicaldisplays143
Principiæ
Structuring thoughts
9
789081
367707
isbn 978 90 813677 07
An engineer fromthe Louvain School of Engineering
and PhDin applied physics fromStanford University,
Jean-luc Doumont nowdevotes his time and energy
to training engineers,scientists,business people,
andother rational minds ineffective communication,
pedagogy,statistical thinking,and related themes.
With his rational background,Jean-luc approaches
communication in an original,engineering-like way
that contrasts sharply with the tradition of the field,
rooted in the humanities.He is thus well received
by students andprofessionals in search of a method
they can apply with the same rigor they have come
to value in every other aspect of their occupations.
An articulate,entertaining,and thought-provoking
speaker,Jean-luc successfully reaches a wide range
of audiences around the world,in English,French,
Dutch,and Spanish—as a trainer or invited speaker
at an array of companies,top-ranked universities,
researchlaboratories,andinternational conferences.
Based on hundreds of interactive training sessions,
this long-awaited book offers Jean-luc’s guidelines
and practical tips toward getting messages across
optimally in written documents,oral presentations,
and graphical displays.Wide-ranging yet compact,
it is true to Jean-luc’s renowned style,proposing
no-nonsense,down-to-earth,readily usable advice
underpinned by a simple yet solid conceptual basis
and presented in a highly modular visual structure.