Sample PPT for testing

1,438 views

Published on

lorem ipsum

Published in: Technology
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,438
On SlideShare
0
From Embeds
0
Number of Embeds
16
Actions
Shares
0
Downloads
0
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Sample PPT for testing

  1. 1. Artificial Intelligence, Cognition and Interaction Kurs: Språk, kommunikation och tänkande Program: Kognitionsvetenskap [email_address]
  2. 2. Topics <ul><li>AI - artificial intelligence </li></ul><ul><ul><li>Cognition </li></ul></ul><ul><ul><li>Language </li></ul></ul><ul><ul><li>Dialogue and Interaction </li></ul></ul><ul><li>AEI - artificial emotional intelligence: </li></ul><ul><ul><li>Emotive agents </li></ul></ul><ul><li>ACI - artificial communicative intelligence </li></ul><ul><ul><li>Non-verbal interactive behavior </li></ul></ul>
  3. 3. Human Brain
  4. 4. Virtual Task Areas <ul><li>Cognition and emotion </li></ul><ul><li>(ICT/ISI Emotion) </li></ul>Speech recognition (IMSC-NL) Integration and Scenario Development (SASO-ST) Human embodiment (ISI-Vhuman) Dialogue management (ICT-NL) Speech understanding and generation (ISI-NL)
  5. 5. AI-DEMO
  6. 6. Kognition <ul><li>Uppmärksamhet - semiotisk organisation av info, färg, proximitet, sekvenser, ordning </li></ul><ul><li>Perception </li></ul><ul><li>Minne </li></ul><ul><ul><li>Recall </li></ul></ul><ul><ul><li>Recognition scan </li></ul></ul><ul><li>Läsa, skriva, tala </li></ul><ul><li>Planera, problemlösning, tänkande, beslutsfattande </li></ul><ul><li>Inlärning - dynalinking </li></ul><ul><li>Känslor </li></ul>
  7. 7. Precortex <ul><li>Används för mål strukturering </li></ul><ul><li>Koordinering av mål och planer </li></ul><ul><li>Förutseende </li></ul>
  8. 8. Inlärning <ul><li>Instruktion </li></ul><ul><li>Immitation - analogi, metafor, exempel </li></ul><ul><li>Upptäckt, sökning </li></ul><ul><li>Längre inlärningstid - stabil inlärning </li></ul><ul><li>Kort inlärningstid - kan glömmas lätt </li></ul>
  9. 9. Mänskliga kriterier för val av problemlösningstillgång <ul><li>Backup avoidance - undvika operatörer som förstör effekten av gammal operatör </li></ul><ul><li>Difference reduction - söka kortaste lösning </li></ul><ul><li>Mean-ends analysis - skapa nytt mål för att kunna tillämpa en operatör, Tower of Hanoi </li></ul>
  10. 10. Mänskliga problemlösnings effekter <ul><li>Funktionell fixering </li></ul><ul><li>Einstellingseffekten, mekaniskt tänkande </li></ul><ul><li>Inkubationseffekten - glöm dåliga lösningar </li></ul><ul><li>Aha effekten - inte medvetna om lösningens närhet </li></ul>
  11. 11. Slutledning och beslutstagande <ul><li>Logik =/ mänskligt tänkade </li></ul><ul><li>Deduktion </li></ul><ul><li>Induktion </li></ul><ul><li>Abduktion (tolking av probabilitet) </li></ul>
  12. 12. Teorier och modeller om känslor <ul><li>Darwinism, biologi, fysiologi </li></ul><ul><li>(William) Jamesianism, psykologi, neurologi </li></ul><ul><li>Kognitivism </li></ul><ul><li>Social konstruktivism, antropologi, sociologi </li></ul>
  13. 13. James, Damasio psykologi, neurologi <ul><li>Upplevelser av känslor </li></ul><ul><li>Kroppens tillstånd påverkar eller orsakar känslan som påverkar tänkandet </li></ul>
  14. 14. Damasio: Descartes' Error <ul><li>Phineas Gage: </li></ul><ul><li>- impaired ability to feel emotion </li></ul><ul><li>- intelligence remained intact after the accident </li></ul><ul><li>- severely handicapped ability to take rational decisions </li></ul><ul><li>Damasio: </li></ul><ul><li>- emotions could no longer be engaged in the decision process </li></ul><ul><li>- rationality stems from our emotions </li></ul><ul><li>- our emotions stem from our bodily senses </li></ul><ul><li>- state of mind is identical to state of feeling, which is a reflection of state of body </li></ul>
  15. 15. Limbic system
  16. 16. Affective Computing Emotionellresponsteknologi <ul><li>Mål </li></ul><ul><ul><li>Att motivera användning, försäljning, inlärning, lek, kreativitet, handling, verksmahet, förtröende </li></ul></ul>
  17. 17. Emotive Agents/Känsloteknologi <ul><li>Frambringa känslor </li></ul><ul><li>Uttrycka känslor </li></ul><ul><ul><li>Virtuella agenter, Pets-Robotar uttrycker känslor </li></ul></ul><ul><ul><li>Användare: Emoticones, emoljud </li></ul></ul><ul><ul><li>Hantera frustrerade användare </li></ul></ul><ul><li>Antropomorfism </li></ul><ul><li>Trovärdighet </li></ul>
  18. 18. ACI <ul><li>Design of virtual agents’ behavior: </li></ul><ul><ul><ul><li>competence </li></ul></ul></ul><ul><ul><ul><li>visualization </li></ul></ul></ul><ul><li>2. Choice of features </li></ul><ul><li>3. Definition of parameters’ values </li></ul><ul><li>4. Evaluation of cultural salience to users </li></ul>
  19. 19. Zero Hypothesis <ul><li>Culture-specific behavior is not detectable via virtual simulations. </li></ul>
  20. 20. Modifiable Computational Model <ul><li>Competence </li></ul><ul><ul><li>Probabilistic Cognitive Models </li></ul></ul><ul><ul><ul><li>Almost all decision involves some chance </li></ul></ul></ul><ul><li>Visualization (ready made) </li></ul><ul><ul><li>Body Movements </li></ul></ul><ul><ul><li>Body Positions </li></ul></ul>
  21. 21. Simulation Setting <ul><li>Outdoor standing groups of 6 men: </li></ul><ul><ul><li>Mexican Spanish speakers </li></ul></ul><ul><ul><li>American English speakers </li></ul></ul><ul><ul><li>Arabic Levantine speakers </li></ul></ul>
  22. 22. Simulated Common Behavior Features <ul><li>Turn-taking (Padilla & Carletta’02) </li></ul><ul><li>Dynamic turn-taking (Jan & Traum’05) </li></ul><ul><li>Movement forces (Jan & Traum’07) </li></ul><ul><li>Gaze (random and probabilistic: listener, addressee, speaker) </li></ul>
  23. 23. <ul><li>talkativeness: likelihood of wanting to talk. </li></ul><ul><li>transparency: likelihood of producing explicit positive and negative feedback, and turn-claiming signals, even gestures. </li></ul><ul><li>confidence: likelihood of interrupting and continuing to speak during simultaneous talk. </li></ul><ul><li>interactivity: the mean length of turn segments between TRPs. </li></ul><ul><li>verbosity: likelihood of continuing the turn after a TRP at which no one is self-selected. </li></ul>Personality Parameters Random choices related to defined parameters
  24. 24. <ul><li>Each reason for movement has a force associated to it (Jan & Traum’07) : </li></ul><ul><ul><li>F speaker : attractive force towards speaker </li></ul></ul><ul><ul><li>F noise : repelling force from outside noise </li></ul></ul><ul><ul><li>F proximity : repelling force from characters that are too close </li></ul></ul><ul><ul><li>F convex : force towards convex hull of all conversation participants </li></ul></ul>Movement Controlling Forces
  25. 25. Simulated Modified Behavior Features <ul><li>Proxemics </li></ul><ul><li>Timing between turns </li></ul><ul><li>Gaze </li></ul>
  26. 26. Literature on Proxemics (eg.Shuter’76) 83cm 79 cm 75 cm 60 cm 56 cm 51 cm 83 cm 71 cm 75 cm 65 cm 50 cm 61 cm M-M M-F F-F M-M M-F F-F Anglo Anglo Anglo Mexican Mexican Mexican Outdoor Adults Indoor Adults Gender Ethnic Groups
  27. 27. Simulated Proxemics (Hall’68) > 2.7 1.2-2.7 0.45-1.2 < 0.45 American English > 1.5 0.7-1.5 0.45-0.7 < 0.45 Arabic Levantine > 2.0 1.0-2.0 < 0.45-1.0 < 0.45 Mexican Spanish Public (m) Social (m) Personal (m) Intimate (m) Proxemic zone/ Ethnic group
  28. 28. Amount of Gaze: Triads and Dyads (Exline’60, Argyle & Ingham’72,’76) 65.7 77.9 47.9 37.9 56.1 73.8 31.1 23.4 37.3 42.4 36.9 7.5 23.2 29.8 25.6 3.0 Mean gaze While listening While talking Mutual gaze Female dyads Male dyads Female triads Male triads Gender/ Gaze (%)
  29. 29. Simulated Gaze Arabic Levantine American English Mexican Spanish Ethnic group 80% 60 % 80% Chance to gaze at speaker
  30. 30. Proxemics <ul><li>Input to movement algorithm </li></ul><ul><li>Force toward/away from agents </li></ul><ul><li>Interacts with other factors such as audibility and group formation to direct agent movement </li></ul>< Proxemics > < IntimateZone >0.45</ IntimateZone > < PersonalZone >1.2</ PersonalZone > < SocialZone >2.7</ SocialZone > </ Proxemics >
  31. 31. Gaze <ul><li>Replace random gaze of original algorithm with gaze based on cultural parameters </li></ul>< Gaze > < GazingAtMeFactor >1.5</ GazingAtMeFactor > < Speaker > < Attending > < Addressee >6.0</ Addressee > < Random >2.0</ Random > < Away >2.0</ Away > </ Attending > < NonAttending > < Addressee >1.0</ Addressee > < Random >8.0</ Random > < Away >1.0</ Away > </ NonAttending > < Away > < Random >9.0</ Random > < Away >1.0</ Away > </ Away > </ Speaker > < Addressee > < Speaker >8.0</ Speaker > < Random >1.0</ Random > < Away >1.0</ Away > </ Addressee > < Listener > < Speaker >6.0</ Speaker > < Addressee >2.0</ Addressee > < Random >1.0</ Random > < Away >1.0</ Away > </ Listener > </ Gaze >
  32. 32. Pause and Overlap <ul><li>Time between speech end and speech start of new speaker in case where the new speaker is taking the turn </li></ul><ul><li>Replace uniform distribution with distribution based on cultural parameters </li></ul>< Silence > < StartOffset >0.0</ StartOffset > < StartVariation >0.5</ StartVariation > </ Silence >
  33. 33. Anglo-American 1
  34. 34. Mexican 1
  35. 35. Arabic 1
  36. 36. Evaluation <ul><li>Subjects evaluate realism in </li></ul><ul><li>6 movies, 2 min each </li></ul><ul><li>20 American speakers </li></ul><ul><li>20 Mexicans speakers </li></ul><ul><li>12 Arabic speakers </li></ul>
  37. 37. Quantitative Results <ul><li>The t-test shows no significant difference in cross-cultural evaluation of gaze and turn-taking </li></ul><ul><li>There are differences in evaluation of proxemics: </li></ul><ul><ul><li>Arabic subjects found Arabic proxemics and animation realistic </li></ul></ul><ul><ul><li>Mexican and Ango-American subjects found no significant cultural differences according to proxemics and overall animation </li></ul></ul><ul><li>Zero hypothesis is violated: </li></ul><ul><ul><li>Subtle culture-specific behavior is detectable via virtual simulations. </li></ul></ul>
  38. 38. Proxemics Ratings <ul><li>Mexicans rate American video 1 as most realistic animation </li></ul><ul><li>Mexicans rate American 2 as least realistic </li></ul><ul><li>Americans thought Mexican 1 is most realistic and Mexican 2 - least realistic </li></ul><ul><li>Americans don't see differences in cultural proxemics </li></ul><ul><li>Arabic speakers rate Arab 1 as most realistic , Arab 2 -least realistic </li></ul>
  39. 39. Promising Integration <ul><li>Cultural awareness can be studied by virtual simulations of interactive behavior </li></ul><ul><li>Simulations trigger research on generalizable parameters and values related to proxemics, gaze and turn-taking </li></ul><ul><li>AEI - triggers development of cognitive-emotion theories </li></ul><ul><li>ACI - triggers development ethics theories of communication </li></ul>

×