the pain of complexity
TRANSCRIPT
The Pain of Complexity
Dmitry ZinovievSuffolk University, Boston
Complexity is one of those fuzzy concepts that everyone seems to understand differently. Other examples: sustainability, resilience, success...
What Is Complexity?
Let's Ask the Mechanical Turk!
Question: Enter three nouns (please, nouns only!) that, to the best of your understanding, are associated with complexity. These words could be synonyms, descriptors, attributes of complexity, actions or any other related words.
Ask 100 mTurk workers, pay a nominal fee for an answer
Reject non-nouns; stem* and aggregate nouns
*We tried both Porter and Lancaster stemmers; Porter is less aggressive, it produced more stems (motifs); some manual tweaking was needed
Direct Survey Results
The stems that were mentioned at least twice (in the order of decreasing frequency):Most frequently mentioned (23+ times): COMPLIC*, DIFFICULTI*, INTRICACI*
Less frequently mentioned (29 times): PROBLEM*, ELABOR*, ENTANGL*, PUZZL*, MULTIPL*, COMPLEX*, COMPUT*, CONVOLUT*, HARD*, RAMIF*, SOPHIST*, CHALLENG*, HUMAN*, INVOLV*, LIFE*, POLIT*, RIDDL*, SCIENC*, COMPOSIT*, CONFUS*, DEPTH*, DESIGN*, DIVERS*, LABYRINTH*, LANGUAG*, MATH*, MAZE*, OBSCUR*, PHYSIC*, QUALITI*, RELATIONSHIP*, SPACE*, STRUCTURE*, TEST*, TRICKI*, TROUBL*, VARIABL*, VARIETI*
The highlighted stems bear strong emotional loadthough, no pain yet!
Use Indirect Data
Don't tell me who you are. Tell me who your friends are, and I will tell you who you are.
Instead of asking people what they think complexity is, let's collect abundant implicit markers off the Internet.
Indirect Data: Possible Sources
EBSCO keywords or subject tags (subscription and harvesting software required; we haz it, but the results may be too academic!)
Blogging sites like LiveJournal (LJ)LJ has individual blogs and community blogs (like forums)
Both types of blogs can and do have user- or moderator-declared interests
The interests are usually carefully chosen to reflect the blogger's/community's online identity
Free access, easy-to-write harvester
Raw Data
All LiveJournal communities that list at least one of the following interests:complexity
complex systems
complexity theory
complexity theories
Found 59 communities, such as mixedtype, abstractthought, ivygreenfanclub, pdmi_logic, investors, gifted_teens, etc.
Many more individual bloggerswe ignored them, their interest lists are way too broad!
Selected 374 most frequently declared interests (the corpus)
Example: Ecologists community
Community ecologists.livejournal.com
150 self-declared (by the moderator(s)) interests
Includes complexity, but also many other keywords; are they relevant? Let's assume they are.
Term Vector Model
Treat communities as documents and interests as words
Calculate generalized similarity between words (and between communitiesbut why?) using Kovacs (2010) algorithm:Two words are similar if they belong to similar documents
Two documents are similar if the consist of similar words
Remark: generalized similarity, as defined by Kovacs, is essentially Pearson correlation in an oblique Cartesian coordinate space
For any two words W and V from the corpus, -1 d(W,V) 1 is the similarity between the words; for any two communities C and D, -1 d(C,D) 1 is the similarity between the communities (a by-product, not needed)
Network Construction (1)
Treat words as nodes in a network
Treat similarities as edges if they are above T0=0.65 (slicing); the choice of T0 is tricky and arbitrary:If T0 is too high, the network is complete and not interesting
If T0 is too low, the network is sparse and disconnected
Gephi it!
Complexity Dragon (a.k.a. the Mindscape of Complexity)
Modular Structure
Modularity=0.56 (not great, but still visible)
psychology, chaos,science, mathematics,complexity theory, ...
philosophy, life,imagination, self-expression,knowledge, ...complexity, honesty, music, simplicity,love, poetry, creativity,empathy, humor, hate, ...writing, art, books,romance, drawing, ...
What Is the Meaning of Modules?
We could just stare at the modules and try to come up with a suitable name
or we could use crowdsourcing through Amazon Mechanical Turk again!
Fragment of the fireball
Let's Ask the Mechanical Turk!
Question: Describe the following group of 50* words with a single most suitable word or a two-word or three-word phrase: ...*There are only 36 terms in the lower left module.
Ask 100 mTurk workers, pay a nominal fee for an answer
Reject non-nouns; stem and aggregate nouns
The nouns are module descriptors (motifs)
Network Construction (2)
Treat original modules as documents and motifs as words
Calculate generalized similarity
Do slicing
Build a network of stems
Gephi it!
The Three Shades of Complexity
The new network has a simple, but not trivial, structure, and is highly modular
The Three Shades of Complexity:Science/Technology
Society/Mind
Creativity/Emotions
Connections:via education
via life/humanities
There are lines here, theyare just too thin...
Semantic Spectrography
mTurk for motifextraction and clustering
Term source formindscape construction
Analysis termselection
Spectrography vs Direct Survey
Direct Survey: 37
Spectrography: 382
4
9
4
104
111
113
37
24
Comparison Summary
Positive:Spectrography is much more detailed: it catches 41% of direct survey terms; direct survey catches only 4.5% of spectrography terms
Spectrography reveals structure
Negative:Spectrography requires a source (or sources) of terms
Successfully Used Elsewhere
D. Zinoviev, D. Stefanescu, L. Swenson, and G. Fireman, Semantic Networks of Interests in Online NSSI* Communities, in Proc. Workshop Words and Networks, Evanston, IL, June 2012, published online (also submitted to Social Networks in 2013)
D.Zinoviev and Z.Zhu, Conceptual Structure of Sustainability: Social and Scholarly Perspectives, Sunbelt XXXIV Social Networks Conference, St.-Pete Beach, FL, February 2014 (also submitted to Social Networks in 2014)
*Non-Suicidal Self Injurya common and epidemically spreading activity among adolescents and young adults
But Where Is the Pain?
SE corner of the emotional module full of NSSI-specific terminology
Communities: artificial_joy, humans_being (sic), the_addicted
Complexity and creativity as attributes of human nature
Thank-You!
Muokkaa otsikon tekstimuotoa napsauttamalla
Muokkaa jsennyksen tekstimuotoa napsauttamallaToinen jsennystasoKolmas jsennystasoNeljs jsennystasoViides jsennystasoKuudes jsennystasoSeitsems jsennystasoKahdeksas jsennystasoYhdekss jsennystaso