550 likes | 679 Views
The Fog of Words. Robert L. Hogenraad Psychology Dept . Université catholique de Louvain, Belgium SDH 2010, Vienna, Oct. 19-20, 2010. robert.hogenraad@uclouvain.be. In social science and the humanities, we do not measure. . Social science.
E N D
The Fog of Words Robert L. Hogenraad PsychologyDept. Université catholique de Louvain, Belgium SDH 2010, Vienna, Oct. 19-20, 2010 robert.hogenraad@uclouvain.be
In social science and the humanities, we do not measure.. Social science We cannot estimate the qualities of « man » using continuous quantitities. However, numbering and classifying qualities by counting is correct for discrete qualities. We count.
9 American-English pan-cultural SD scales About questionnaires « That’s not the way people talk » (a remark by Charlie Osgood) nice :___:___:___:___:___:___:___: awful powerless :___:___:___:___:___:___:___: powerful fast :___:___:___:___:___:___:___: slow bad :___:___:___:___:___:___:___: good big :___:___:___:___:___:___:___: little dead :___:___:___:___:___:___:___: alive sweet :___:___:___:___:___:___:___: sour weak :___:___:___:___:___:___:___: strong young :___:___:___:___:___:___:___: old
9 pan-cultural RussianSD scales N@D@T46 :___:___:___:___:___:___:___: B:@N@6 F:"$Z6 :___:___:___:___:___:___:___: F4:\>Z6 $ZFHDZ6 :___:___:___:___:___:___:___: <,*:,>>Z6 >,BD4bH>Z6 :___:___:___:___:___:___:___: BD4bH>Z6 $@:\T@6 :___:___:___:___:___:___:___: <":,>\846 FB@8@6>Z6 :___:___:___:___:___:___:___: &@2$J0*.>>Z6 BD,8D"F>Z6 :___:___:___:___:___:___:___: J0"F>Z6 :.(846 :___:___:___:___:___:___:___: Hb0.:Z6 >@&Z6 :___:___:___:___:___:___:___: FH"DZ6
Ruleof criticalasymmetry: • « The more clear and transparent a text, the less effort isrequired of the reader. » AMD The hermeneutic chiasma
The less structured a text, the more structured and categorizing the analysis must be. • Example: the postcard that lands by mistake in your mailbox… AMD In content analysis…
AMD… The mystery postcard …
A: “Bill, you’d better get that Linden back or you’ll lose that baby too.” • B: “Yeah, I just lost 81.” • B: “Look any better?” • A: “No. You still got to get rid of about 400 Bill because you’re 400 over the short time emergency on that 80 line.” • B: “Yeah - that’s what I’m saying. Can you help me with that?” Technical conversations between experts … … turn easily into a quasi-private language because so many elements of it are implicit, i.e., the text is structured only for the experts, but not for outsiders.
…conversations between Senior Pool Dispatcher (A) and Con Edison System Operator (B) between 8:56pm and 9:02pm, July 13, 1977. • (Extracts from the State of New York Investigation, New York City Blackout, July 13, 1977, p. 13). Technical conversations .between experts …
Content Analysis: An Emerging Zone Epistemologicalissues Statistical issues Content analysis to the service of Policy planning
Two traditions for analyzing texts, patristic and rabbinic, or nomothetic and idiothetic. • And when to use each of them Spicing up content analysis with epistemology
« It is He Who sent down the Book upon you. In it are verses precise in meaning… Others are ambiguous. Those in whose heart is waywardness pursue what is ambiguous therein, seeking discord… » (The House of ‘Imran, 3:7). The two traditions of content analysis in The QUR’AN The Qur’an. A new translation by Tarif Khalidi
(1. Epistemological issues) The nomothetic tradition in content analysis. example
SCORES FOR CATEGORY BASED DICTIONARY : CAT. 33 N-Power :..............20-30 SEG TEXT WORDS FREQ SQRT RATE DENS SQRT RATE TOTAL DIFF. FREQ DENS __________________________________________________________________ 13 4414 1011 351 8.917 135 5.530 14 13072 2366 1076 9.073 367 5.299 15 4482 1129 405 9.506 162 6.012 16 21026 2593 1502 8.452 362 4.149 17 13571 2041 1155 9.225 297 4.678 18 15102 2350 1076 8.441 329 4.667 19 7736 1437 486 7.926 166 4.632 ……………
Risk of conflict = nPow minus nAff -- in D. C. McClelland’s motive theory.
(1. Epistemological issues) The idiothetic tradition in content analysis. example
Word-word correlations (fictive) JOHN FATHER TRUE PEACE WAR FIGHT POLITICS BIRD MARY JOHN 1,0000 ,1021 ,1666 -,1554 ,2015 ,0552 -,3038 -,3104 -,2303 FATHER ,1021 1,0000 -,0422 -,0595 ,4005 -,3870 -,2054-,3768 ,1373 TRUE ,1666 -,0422 1,0000 -,3055 ,2247 ,3311 -,2182-,2237 -,0776 PEACE -,1554 -,0595 -,3055 1,0000 -,5062 -,5301 ,9704*** -,0680 ,3822 WAR ,2015 ,4005 ,2247 -,5062 1,0000 ,1889 -,5650-,2206 -,1330 FIGHT ,0552 -,3870 ,3311 -,5301 ,1889 1,0000 -,4083,5007 -8396** POLITICS -,3038 -,2054 -,2182 ,9704*** -,5650 -,4083 1,0000,0011 ,3453 BIRD -,3104 -,3768 -,2237 -,0680 -,2206 ,5007 ,00111,0000 -,4657 MARY -,2303 ,1373 -,0776 ,3822 -,1330 -,8396** ,3453-,4657 1,0000 Probability 2-tails : * - .05 ** - .01 ***. -.001
…and joint correspondence analysis (over 9 words and 10 observations) …
RETAKE Comparing the nomothetic & the idiothetic traditions in content analysis
observer instrument (telescope) object (moon) text analyst dictionary (marker) text
…is ordinary in kind, • but allows one to reach extraordinary outcome in degree. The patristic –dictionary– tradition to analyze texts …
What the hell is water? In the rabbinic tradition, no instrument between analyst and text Analyst and text are in the same glass
The idiothetic« rabbinic » tradition In The QUR’AN…
The two traditions (a) • Trad 2: Looks for clusters thatmayrefer to a theme** **Cluster of wordswithdifferentmeanings • Trad 1: Forcefully substitutes words of a text with categories* (=dictionaries) • *Group of words with similar meanings
The two traditions (b) • Trad 2: Yields complex themes from fragments of a text, yet no unique solution • Trad 1: Dictionaries are calibrated instruments, leave no space for doubt
The two traditions (c) • Trad 2: Words as symptoms* • * About unverifiableinterpretations:It is easy for human observers to see the response they want and so to be fooled by the data • Trad 1: Words as predictors
The two traditions (d) • Trad 2: One looks for contiguitiesbetweenwords of the text in order to discover latent meanings #commenting the text, withoutalteringit • Trad 1: Tradition of distrust #answers to pre-existing questions [« You’veleft out everythingwhichdoesn’t fit », in Tom Stoppard’sArcadia, p. 59]
The two traditions (e) • Trad 2: Idiothetic, contiguities are unique, never seen before, never to be seen after • Trad 1: Dictionariessharefeatureswithothertexts
It is inefficient to attempt to analyze complex textual data using a complex interpretative tool … • … as in this … Retake
When the General in charge of the Afghanistan war saw this graph, he said: • «When we’ll have understood that, we’ll have won the war!»
The heart of the argument is that one cannot analyze sampling error in the case of a unique narrative event. History happens only once (1)
The Lloyd’s of London have an “Unusual Risks” section, because there is no distribution theory for unique events to turn to. History happens only once (2)
The once-ness of every inference model Troy Polamalu is an American football player (from Samoa). He didn’t cut his hair for the last 7 years. They say that one hair after another, his hair is 11 kilometers long! That wouldn’t happen to me…
The once-ness of every inference model Now, the Head & Shoulder cosmetic company for which he makes ads insured his hair with the Lloyds of London for 1 million $.
Rate of primordial thought contents (RID) in Ulysses’ 18 chapters
Rate of primordial thought contents in Ulysses’ 18 chapters after 10 bootstrap simulations
At the end of the day, whether or not one agrees with the conclusions is less important than the insight one can gain from recognizing the importance of the rule of critical asymmetry. Afterword And then … “Das LebenGehtWeiter”
3. To the service of policy planning Usingnumbers to predictbehavior
Predicting the risk of war in the speeches of President Medvedev(January 24 to September 11, 2008) Content analysis to the service of predicting conflicts Dictionary is the Motive Dictionary (version 6.0)
When you deal with narrativity, you need to use statistics that are sensitive to it, like change-point tests Where are the cut-off points in a text? Answer: Use CART (for Classification and Regression Trees)