lecture 6: multitasking › ~salvucci › courses › cs630... · the multitasking continuum...

Post on 30-Jun-2020

10 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Page 1

1 CS 630: Cognitive Systems. Dario Salvucci, Drexel University.

Lecture 6: Multitasking

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 2

Our Multitasking World

talking & driving

cooking & reading a book

writing paper & reading email

watching game & talking to friends

listening & note-taking

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 3

The Challenges of Multitasking

■  Technological challenges –  user interfaces, hardware, networking...

■  Scientific challenges –  how do we multitask? –  when is multitasking easy or difficult? –  how does it affect task performance?

■  Societal challenges –  when is multitasking useful? –  when is multitasking inappropriate? –  when is multitasking dangerous?

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 4

The Multitasking Continuum

talking & driving

cooking & reading a book

writing paper & reading email

watching game & talking to friends

listening & note-taking

Page 2

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 5

The Multitasking Continuum

Concurrent Multitasking (e.g., PRP, driver distraction)

Sequential Multitasking (e.g., task interruptions)

seconds hours minutes

Time between Task Switches

talking & driving

cooking & reading a book

writing paper & reading email

watching game & talking to friends

listening & note-taking

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 6

Threaded Cognition

■  Goal: Unifying theory of multitasking –  ... across the multitasking continuum –  ... across laboratory and real-world domains –  ... across different levels of abstraction

■  Approach: Computational cognitive modeling (obviously J) –  Threaded cognition –  in the ACT-R cognitive architecture

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 7

Threaded Cognition

■  Your brain is a “Thought Kitchen” –  with resources and processes

•  central resource: the cook •  other resources: oven, stove, mixer, etc.

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 8

Threaded Cognition

■  Concurrent multitasking is a basic skill—best represented by a simple general mechanism

■  Threaded cognition... –  allows concurrent execution of multiple “streams of thought” = threads

–  takes models A, B... predicts behavior of A+B

■  Theoretical components –  (1) Resources that perform relevant processing

•  derived from the ACT-R architecture

–  (2) Processing principles that define task allocation

Page 3

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 9

Processing Principles

(1) Threaded Processing –  Cognition maintains a set of active goals that

produce “threads” of processing •  many domains are nicely represented as threads

–  some are more obvious – e.g., driving + dialing –  some are less obvious – e.g., list-memory tasks

–  In ACT-R terms, this means maintaining multiple goals at a time

•  In the past, ACT-R had only one goal at a time •  then it had a “goal stack” (inspired by tasks with a

robust subgoal structure, like Tower of Hanoi) •  now, several active goals

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 10

Processing Principles

(2) Resource Exclusivity –  Resources execute processing requests serially,

exclusively for one request/task at a time •  resources can be massively parallel themselves,

within the resource –  e.g., visual processing

•  but resources can serve only one goal at a time

–  (caveat: what about resources like motor? — are the hands independent? fingers? hands from feet? etc.)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 11

Processing Principles

(3) Resource Usage –  Threads acquire and release resources in a

greedy, polite manner. •  greedy: used as soon as available •  polite: threads free resources ASAP

(4) Conflict Resolution –  When threads contend for the procedural

resource, the thread with the highest urgency proceeds.

•  highest urgency = least recently used •  simple mechanism for balancing thread processing

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 12

Threaded Models

■  Now let’s look at some models that use threaded cognition…

Page 4

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 13

Threads-1

(set-­‐task  "cs630.ThreadsTask”    (add-­‐dm          (type-­‐number              isa  type-­‐number)          (answer-­‐phone              isa  answer-­‐phone)    (goal-­‐focus  type-­‐number)  (goal-­‐focus  answer-­‐phone)  

(p  type-­‐number*Bind-­‐Birst          =goal>                  isa  type-­‐number                  current-­‐x  nil          ?visual-­‐location>                  state  free                  buffer  empty        ?visual>                  state  free                  buffer  empty    ==>          +visual-­‐location>                  isa  visual-­‐location                  screen-­‐x  lowest  )  

(p  type-­‐number*Bind-­‐next          =goal>                  isa  type-­‐number                  current-­‐x  =x          ?visual-­‐location>                  state  free                  buffer  empty          ?visual>                  state  free                  buffer  empty    ==>          +visual-­‐location>                  isa  visual-­‐location                  screen-­‐x  lowest                  >  screen-­‐x  =x  )  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 14

Threads-1

(p  type-­‐number*encode          =goal>                  isa  type-­‐number          =visual-­‐location>                  isa  visual-­‐location                  screen-­‐x  =x          ?visual>                  state  free                  buffer  empty  ==>          +visual>                  isa  move-­‐attention                  screen-­‐pos  =visual-­‐location          =goal>                  current-­‐x  =x  )  

(p  type-­‐number*type          =goal>                  isa  type-­‐number          =visual>                  isa  text                  value  =digit          ?manual>                  state  free  ==>          +manual>                  isa  press-­‐key                  key  =digit  )  

(p  type-­‐number*done          =goal>                  isa  type-­‐number                  current-­‐x  =x          ?visual-­‐location>                  state  error  ==>          -­‐goal>  )  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 15

Threads-1

(p  answer-­‐phone*encode          =goal>                  isa  answer-­‐phone          =aural-­‐location>                  isa  audio-­‐event          ?aural>                  buffer  empty                  state  free  ==>          +aural>                  isa  ring                  event  =aural-­‐location  )  

(p  answer-­‐phone*done          =goal>                  isa  answer-­‐phone          =aural>                  isa  ring          ?vocal>                  state  free  ==>          +vocal>                  isa  speak                  string  "Hello!"          -­‐goal>  )  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 16

Threads-1 0.000 vision unrequested [vision~62]! 0.000 procedural start! 0.050 procedural ** TYPE-NUMBER*ENCODE ** [type-number]! 0.050 vision move-attention! 0.135 vision encoding-complete [text~66]! 0.185 procedural ** TYPE-NUMBER*TYPE ** [type-number]! 0.185 motor press-key "1"! 0.235 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number]! 0.235 vision find-location [vision~72]! 0.285 procedural ** TYPE-NUMBER*ENCODE ** [type-number]! 0.285 vision move-attention! 0.300 audio audio-event [audio-event~63]! 0.300 audio unrequested [audio-event~63]! 0.350 procedural ** ANSWER-PHONE*ENCODE-SOUND ** [answer-phone]! 0.350 audio attend-sound! 0.370 vision encoding-complete [text~75]! 0.435 motor preparation-complete! 0.485 motor initiation-complete! 0.585 motor output key 1! 0.735 motor finish-movement! 0.785 procedural ** TYPE-NUMBER*TYPE ** [type-number]! 0.785 motor press-key "2"! 0.835 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number]! 0.835 vision find-location [vision~84]! 0.850 audio audio-encoding-complete [ring~78]! 0.885 procedural ** TYPE-NUMBER*ENCODE ** [type-number]! 0.885 vision move-attention! 0.935 motor preparation-complete!

Page 5

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 17

Threads-1 0.935 procedural ** ANSWER-PHONE*DONE ** [answer-phone]! 0.935 declarative store chunk [answer-phone]! 0.935 speech speak "Hello!"! 0.970 vision encoding-complete [text~87]! 0.985 motor initiation-complete! 1.085 motor output key 2! 1.085 speech preparation-complete! 1.135 speech initiation-complete! 1.135 speech output-speech "Hello!"! 1.235 motor finish-movement! 1.285 procedural ** TYPE-NUMBER*TYPE **! 1.285 motor press-key "3"! 1.335 procedural ** TYPE-NUMBER*FIND-NEXT **! 1.335 vision find-location [vision~95]! 1.385 procedural ** TYPE-NUMBER*ENCODE **! 1.385 vision move-attention! 1.435 speech finish-movement! 1.435 motor preparation-complete! 1.470 vision encoding-complete [text~98]! 1.485 motor initiation-complete! 1.585 motor output key 3! 1.735 motor finish-movement! 1.785 procedural ** TYPE-NUMBER*TYPE **!...!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 18

Threads-2

(set-­‐task  "cs630.ThreadsTask")    (add-­‐dm          (start  isa  goal)  )    (goal-­‐focus  start)  

(p  start-­‐multitasking          =goal>                  isa  goal  ==>          +goal>                  isa  type-­‐number          +goal>                  isa  answer-­‐phone  )  

...    <  same  as  before>  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 19

Threads-2 0.000 vision unrequested [vision~62]! 0.000 procedural start! 0.050 procedural ** START-MULTITASKING **! 0.050 declarative store chunk [start] (start isa goal)! 0.050 declarative store chunk [type-number~65]! 0.100 procedural ** TYPE-NUMBER*ENCODE ** [type-number~65]! 0.100 vision move-attention! 0.185 vision encoding-complete [text~70]! 0.235 procedural ** TYPE-NUMBER*TYPE ** [type-number~65]! 0.235 motor press-key "1"! 0.285 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number~65]! 0.285 vision find-location [vision~76]! 0.300 audio audio-event [audio-event~63]! 0.300 audio unrequested [audio-event~63]! 0.335 procedural ** TYPE-NUMBER*ENCODE ** [type-number~65]! 0.335 vision move-attention! 0.385 procedural ** ANSWER-PHONE*ENCODE-SOUND ** [answer-phone~67]! 0.385 audio attend-sound!...! 0.835 procedural ** TYPE-NUMBER*TYPE ** [type-number~65]! 0.835 motor press-key "2"! 0.885 audio audio-encoding-complete [ring~82]! 0.885 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number~65]! 0.885 vision find-location [vision~88]! 0.935 procedural ** ANSWER-PHONE*DONE ** [answer-phone~67]! 0.935 declarative store chunk [answer-phone~67]! 0.935 speech speak "Hello!"!...!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 20

Threads-3

(set-­‐task  "cs630.ThreadsTask")    (add-­‐dm          (type-­‐number              isa  type-­‐number)  )    (goal-­‐focus  type-­‐number)    ...  

(p  handle-­‐sound*encode          =goal>          =aural-­‐location>                  isa  audio-­‐event          ?aural>                  buffer  empty                  state  free  ==>          +aural>                  isa  ring                  event  =aural-­‐location          +goal>                  isa  handle-­‐sound          +goal>  =goal  )  

(p  handle-­‐sound*phone          =goal>                  isa  handle-­‐sound          =aural>                  isa  ring          ?vocal>                  state  free  ==>          +vocal>                  isa  speak                  string  "Hello!"          +goal>                  isa  converse  )  

Page 6

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 21

Threads-3 0.000 vision unrequested [vision~62]! 0.000 procedural start! 0.050 procedural ** TYPE-NUMBER*ENCODE **! 0.050 vision move-attention! 0.135 vision encoding-complete [text~66]! 0.185 procedural ** TYPE-NUMBER*TYPE **! 0.185 motor press-key "1"! 0.235 procedural ** TYPE-NUMBER*FIND-NEXT **! 0.235 vision find-location [vision~72]! 0.285 procedural ** TYPE-NUMBER*ENCODE **! 0.285 vision move-attention! 0.300 audio audio-event [audio-event~63]! 0.300 audio unrequested [audio-event~63]! 0.350 procedural ** HANDLE-SOUND*ENCODE **! 0.350 declarative store chunk [type-number]! 0.350 audio attend-sound! 0.370 vision encoding-complete [text~75]! 0.435 motor preparation-complete! 0.485 motor initiation-complete! 0.585 motor output key 1! 0.735 motor finish-movement! 0.785 procedural ** TYPE-NUMBER*TYPE ** [type-number~80]! 0.785 motor press-key "2"! 0.835 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number~80]! 0.835 vision find-location [vision~87]! 0.850 audio audio-encoding-complete [ring~81]! 0.885 procedural ** TYPE-NUMBER*ENCODE ** [type-number~80]! 0.885 vision move-attention!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 22

Threads-3 0.935 motor preparation-complete! 0.935 procedural ** HANDLE-SOUND*PHONE ** [handle-sound~79]! 0.935 declarative store chunk [handle-sound~79]! 0.935 speech speak "Hello!"! 0.970 vision encoding-complete [text~90]! 0.985 motor initiation-complete! 1.085 motor output key 2! 1.085 speech preparation-complete! 1.135 speech initiation-complete! 1.135 speech output-speech "Hello!"! 1.235 motor finish-movement! 1.285 procedural ** TYPE-NUMBER*TYPE ** [type-number~80]! 1.285 motor press-key "3"! 1.335 procedural ** TYPE-NUMBER*FIND-NEXT ** [type-number~80]! 1.335 vision find-location [vision~100]! 1.385 procedural ** TYPE-NUMBER*ENCODE ** [type-number~80]! 1.385 vision move-attention! 1.435 speech finish-movement! 1.435 motor preparation-complete! 1.470 vision encoding-complete [text~103]! 1.485 motor initiation-complete! 1.585 motor output key 3! 1.735 motor finish-movement! 1.785 procedural ** TYPE-NUMBER*TYPE ** [type-number~80]! 1.785 motor press-key "4"!!< “converse” goal is still active and proceeds here... >!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 23

To the Laboratory...

■  Let’s look at threaded cognition in two laboratory tasks: –  tracking & choice –  dual-choice tasks

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 24

Tracking & Choice

■  Manual tracking appears in many forms

Page 7

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 25

Tracking

■  In many experiments, it’s much more controlled –  either: try to keep a pointer on a target

–  or: try to keep a cursor within a target range

–  while the movement is generated using a pseudo-random forcing function

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 26

Tracking & Choice

■  Experiment: Martin-Emerson & Wickens (1992) –  tracking: keep the cursor in

the target area •  hard vs. easy tracking,

depending on forcing function

–  choice: see an arrow pointing {left, right}, press key to respond

•  arrow separated from target area by offset that varies between 0° and 35° of visual angle

target area

cursor

choice stimulus

offset

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 27

Tracking & Choice

■  Experiment: Martin-Emerson & Wickens (1992)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 28

Track.java

public void start (){ processDisplay(); Utilities.shuffle (offsetIndices); offsetIndex = 0; offsetCount = 0; lastArrowTime = 5.0;

addPeriodicUpdate (.020);}

void updateTarget (double time){ double pi2 = 2 * Math.PI; if (easy) { tx = (2.86 * Math.sin (0.00 + (pi2 * (time / 16.670)))) + (1.15 * Math.sin (1.57 + (pi2 * (time / 6.250)))) + (0.57 * Math.sin (3.93 + (pi2 * (time / 9.091))));

ty = (2.29 * Math.sin (0.79 + (pi2 * (time / 8.000)))) + (1.72 * Math.sin (4.72 + (pi2 * (time / 11.110)))) + (1.72 * Math.sin (2.36 + (pi2 * (time / 50.000)))); } else ...

<< move target to (tx,ty) >>}

Page 8

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 29

Track (set-­‐task  "cs630.Tracking")    (sgp          :v  nil          :emma  t  )    (start-­‐hand-­‐at-­‐mouse)    (add-­‐dm          (track-­‐goal  isa  track)          (choice-­‐goal  isa  choice)  )    (goal-­‐focus  track-­‐goal)  (goal-­‐focus  choice-­‐goal)  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 30

Track ;;  Tracking  Task    (p  Bind-­‐target          =goal>                  isa  track          ?visual-­‐location>                  state  free                  -­‐  buffer  requested          ?visual>                  state  free                  buffer  empty  ==>          +visual-­‐location>                  isa  visual-­‐location                  kind  cross  )  

(p  move-­‐to-­‐target          =goal>                  isa  track          =visual-­‐location>                  kind  cross          ?visual>                  state  free                  buffer  empty          ?manual>                  state  free  ==>          +visual>                  isa  move-­‐attention                    screen-­‐pos  =visual-­‐location          +manual>                  isa  move-­‐cursor                  loc  =visual-­‐location  )  

(p  repeat-­‐track          =goal>                  isa  track          =visual>                  isa  cross          ?manual>                  state  free  ==>  )  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 31

Track ;;  Choice  Task    (p  Bind-­‐arrow          =goal>                  isa  choice          ?visual-­‐location>                  state  free                  -­‐  buffer  requested          ?visual>                  state  free                  buffer  empty  ==>          +visual-­‐location>                  isa  visual-­‐location                  kind  text                  :attended  nil  )  

(p  arrow-­‐not-­‐found          =goal>                  isa  choice          ?visual-­‐location>                  state  error  ==>          -­‐visual-­‐location>  )    (p  encode-­‐arrow          =goal>          =visual-­‐location>                  kind  text          ?visual>                  state  free                  buffer  empty  ==>          +visual>                  isa  move-­‐attention                    screen-­‐pos  =visual-­‐location  )  

(p  respond-­‐left          =goal>                  isa  choice          =visual>                  isa  text                  value  "<"          ?manual>                  state  free  ==>          +manual>                  isa  punch                  hand  left                  Binger  pinkie  )    (p  respond-­‐right        ...  )  

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 32

Track 1849.957 procedural ** FIND-ARROW ** [choice-goal]! 1849.957 vision error! 1850.007 procedural ** ARROW-NOT-FOUND ** [choice-goal]! 1850.020 vision unrequested [vision~27824]! 1850.057 procedural ** FIND-TARGET ** [track-goal]! 1850.057 vision find-location [vision~27827]! 1850.107 procedural ** MOVE-TO-TARGET ** [track-goal]! 1850.107 vision move-attention! 1850.107 motor move-cursor vision~27827! 1850.107 motor preparation-complete! 1850.148 vision encoding-complete [cross~27832]! 1850.157 motor initiation-complete! 1850.242 eye preparation-complete [cross~27832]! 1850.315 eye execution-complete [cross~27832]! 1850.529 motor move cursor (255 10)! 1850.579 motor finish-movement! 1850.629 procedural ** REPEAT-TRACK ** [track-goal]! 1850.679 procedural ** FIND-ARROW ** [choice-goal]! 1850.679 vision find-location [vision~27836]! 1850.729 procedural ** ENCODE-ARROW ** [track-goal]! 1850.729 vision move-attention! 1850.864 eye preparation-complete [text~27839]! 1850.995 eye execution-complete [text~27839]! 1851.023 vision encoding-complete [text~27839]! 1851.073 procedural ** RESPOND-LEFT ** [choice-goal]! 1851.073 motor punch left pinkie! 1851.123 procedural ** FIND-TARGET ** [track-goal]! ...!

Page 9

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 33

Track

■  Process timeline

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 34

Tracking & Choice

■  Results: Choice response time

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 35

Tracking & Choice

■  Results: Tracking error

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 36

Dual-Choice Tasks

■  A choice task means –  get a (simple) stimulus –  produce a (simple) response

■  Dual-choice tasks ask a person to do 2 choice tasks at almost the same time

■  Several factors are often varied in experiments using this paradigm –  perceptual modality: visual / aural –  motor modality: manual / vocal –  cognitive difficulty of stimulus à response

mapping

Page 10

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 37

Dual-Choice Tasks

■  Example (with consistent SàR mappings) –  visual-manual task

–  aural-vocal task

O – – – – O – – – – O –

low tone

one

mid tone

two

high tone

three

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 38

Dual-Choice Tasks

■  In dual-choice tasks, there’s usually a delay between one stimulus and the other –  called the “SOA” = “stimulus onset asynchrony” –  e.g., if we start the aural-vocal task at time = 0,

we might present the visual-manual task at time = {.000 .050 .150 .250 .500 1.000}

■  Compared to the single-task case, how long will the visual-manual task take when... –  SOA is big? (stimuli far apart)

–  SOA = 0? (concurrent stimuli)

–  somewhere in between?

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 39

The PRP Effect

■  Psychological Refractory Period (PRP) effect –  “refractory” from analogy with cells returning

to normal after excitation (not a great analogy, but it stuck)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 40

The PRP Effect

■  What causes the PRP effect? ■  For a long time, it was assumed to be a solid

indicator of a “cognitive bottleneck” ■  A box-diagram depiction:

–  this is the “response-selection” bottleneck –  it’s also a bit misleading...

Is there really an inherent cognitive bottleneck?

Page 11

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 41

Experiments

■  Schumacher et al. (2001), Experiment 1 –  people can achieve perfect time-sharing!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 42

Experiments

■  Schumacher et al. (2001), Experiment 1 –  people can achieve perfect time-sharing! –  table with final results after learning only...

Single-Task Dual-Task

Aural-Vocal 446 ms 456 ms

Visual-Manual 281 ms 283 ms

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 43

Experiments

■  Schumacher et al. (2001), Experiment 2 –  PRP effect comes from instructions /

constraints: do Task 1, then do Task 2

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 44

Perfect Time-Sharing

■  Perfect time-sharing follows readily from threaded cognition –  in this case, everything works out perfectly; no

interference, no dual-task/PRP effect!

Page 12

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 45

Dual-Choice Tasks

■  An inconsistent SàR mapping requires an extra step to retrieve the mapping

O – – – – – – O

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 46

Dual-Choice Tasks

■  An inconsistent SàR mapping requires an extra step to retrieve the mapping –  with possible interference, as here (A)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 47

Dual-Choice Tasks

■  Increased perceptual difficulty makes perception for the tasks run into each other

O O O O O O O O

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 48

Dual-Choice Tasks

■  Increased perceptual difficulty makes perception for the tasks run into each other –  with possible interference, as here (B)

Page 13

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 49

Problem State

■  Problem state = temporary information required during task execution –  roughly speaking, a task’s “mental context” –  in ACT-R: stored in the imaginal buffer

■  Example: Solving 3+4 –  encode “3”, then “+”, then “4” –  all this is now held in the problem state /

imaginal buffer –  in this case, used to pass along information for

retrieval –  can also be used to remember new information

•  i.e., associate 3, +, 4, and then 7 with one another

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 50

Problem State

■  Example: Writing a paper –  what do you need to keep in mind as you’re

writing a... •  sentence? •  paragraph? •  section?

–  where do you maintain the least amount of information?

■  Example: Tracking, or Driving –  no problem state needed!

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 51

Memory & Interruptions

■  Memory clearly plays an important role in interruptions

■  What are the (at least two) important features of human memory? –  information strengthens with use –  information decays over time

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 52

Memory for Goals

■  Memory for goals theory (Altmann & Trafton, 2002) + ACT-R memory theory (Anderson et al., 2004) –  to suspend a task, people encode (rehearse) the

current goal until it’s readily available in memory •  in ACT-R, each retrieval boosts a chunk’s activation,

making it easier to recall •  e.g., rehearse “I’m ordering a platypus” a few times

–  to resume the task, people simply recall the goal •  in ACT-R, associated cues can facilitate recall •  e.g., seeing computer, or browser on platypus web page

Page 14

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 53

Memory for Goals

■  Memory for goals as threads...

■  Encoding

■  Retrieval

Primary task Secondary task Primary task

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 54

Memory for Goals

■  Memory for goals as threads...

■  Encoding

■  Retrieval

Rehearsal Retrieval Primary task

Secondary task

Primary task

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 55

Memory for Goals

■  How many retrievals/rehearsals is a “good” number for a typical interruption?

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 56

Interruption Study

■  Monk, Trafton, & Boehm-Davis (2008) –  explored effects of interruption duration &

demand on primary-task resumption

–  primary task: programming a VCR –  interruption duration: 3, 8, or 13 seconds –  interrupting task

•  no-task: just wait •  track: manual tracking task •  n-back: compare current and previous letters (<,>)

Page 15

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 57

Interruption Study

■  Model –  model for primary task

•  full model not needed… •  specified declarative chunk to represent the goal •  estimated time parameter for performing first action

–  models for interrupting tasks •  no-task: trivially waits •  tracking: does the tracking

•  n-back: simplified from previous work (Juvina & Taatgen, 2007)

–  uses declarative resource to retrieve last item!

–  model for interruption process described earlier –  but when exactly should encoding occur?

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 58

Interruption Study

■  Encoding strategies –  S1: Encode during the entire interruption –  S2: Encode for n seconds, concurrently with

secondary task –  S3: Encode until retrieval takes no more than n

seconds, concurrently with the secondary task –  S4: Encode for n seconds prior to the

interruption, ending at the onset of the interruption

–  S5: Encode for a few (3) retrievals prior to the interruption, ending at the onset of the interruption

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 59

Interruption Study

Monk et al. (2008)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 60

Interruption Study

Monk et al. (2008) Model – S1

(rehearse entire interruption)

no effect of duration

Page 16

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 61

Interruption Study

Monk et al. (2008) Model – S5

(rehearse few times before interrupt)

duration effect too large; no n-back interaction

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 62

Interruption Study

Monk et al. (2008) Model – S4

(rehearse for n sec before interrupt)

no n-back interaction

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 63

Interruption Study

Monk et al. (2008) Model – S3

(rehearse until retrieval < n sec)

strange n-back interaction (interference forces���too much encoding!)

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 64

Interruption Study

Monk et al. (2008) Model – S2

(rehearse for n sec after interrupt)

Yes! n-back interaction due to���declarative interference

R2=.94

Page 17

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 65

Interruption Study

■  ACT-R model for S2 –  interleaved with tracking

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 66

Interruption Study

■  ACT-R model for S2 –  interleaved with the N-back task

CS 630: Cognitive Systems. Dario Salvucci, Drexel University. 67

Interruption Study

■  Tracking error –  data: slight effect for 3-sec interruption across

three experiments (albeit not conclusive) –  model: S2 & S3 show this effect due to encoding

■  Bottom line: Memory-intensive interruptions (like N-back) are especially disruptive because they interfere with rehearsal

top related