IONS-VIP: A cognitive model for navigating the web via - ACT-R

Modeling the N-back-M-pitch paradigm
Ion Juvina*, Michael Qin^, & Christian Lebiere*
*Department of Psychology, Carnegie Mellon University
^Naval Submarine Medical Research Laboratory
1
Task
|
Study
|
Model
|
Discussion
|
Future
Outline

Task

Study

Cognitive model

Discussion

Future work
2
Task
|
Study
|
Model
|
Discussion
|
Future
Diving

Breathing compressed air at depth

Hyperbaric nitrogen narcosis



lightheadedness, inattention, difficulty
concentrating, poor judgment, decreased
coordination
Resembles ethanol intoxication
Becomes noticeable at a depth of ~130 fsw (~5
ATA)
3
Task
|
Study
|
Model
|
Discussion
|
Future
Performance impairment
4
Task
|
Study
|
Model
|
Discussion
|
Future
N-Back

Auditory N-back


Acute effects of ethanol
N-back M-Pitch (Qin et al., 2011)


Same / different duration as 1-back
Added pitch as a distracting feature

Pitch variation





Standard pitch: 700 Hz, p = 0.7
Slightly deviant: 750 Hz, p = 0.1
Moderately deviant: 900 Hz, p = 0.1
Widely deviant: 1200 Hz, p = 0.1
ISI = ~ 1s
5
Task
|
Study
|
Model
|
Discussion
|
Future
N-Back M-Pitch
6
Task
|
Study
|
Model
|
Discussion
|
Future
Method

Participants:


Apparatus:


5 US Navy trained divers
Hyperbaric chambers to simulate diving
Conditions:

Pre-dive, Dive at 190 fsw, Post-dive

Pre-dive, Dive at 280 fsw, Post-dive
7
Task
|
Study
|
Model
|
Discussion
|
Future
Method
8
Task
|
Study
|
Model
|
Discussion
|
Future
Results

Performance decreases with dive depth

The cognitive deficit associated with
narcosis is temporary
9
Task
|
Study
|
Model
|
Discussion
|
Future
Accuracy at 190 fsw
10
Task
|
Study
|
Model
|
Discussion
|
Future
Accuracy at 280 fsw
11
Task
|
Study
|
Model
|
Discussion
|
Future
Relative accuracy
12
Task
|
Study
|
Model
|
Discussion
|
Future
Model assumptions

Main challenge of the task:

Maintaining focal attention in the face of
distraction

Distraction caused by low frequency of
deviant pitches

Diving reduces ability to deal with
interference
13
Task
|
Study
|
Model
|
Discussion
|
Future
Model description

Correct performance





Errors


Perceive current sound, encode sound length
Retrieve previous sound
Compare perceived and retrieved sounds
Respond same / different
Retrieval of previous sound perturbed by distracting feature of
stimulus: partial matching
Performance deterioration at depth

Parameters



Pre- and post-dive: ans=0.1 ; mp = 1.0
Dive at 190 fsw: ans = 0.2 ; mp = 1.5
Dive at 280 fsw: ans = 0.25 ; mp = 2.0
14
Task
|
Study
|
Model
|
Discussion
|
Future
Correct performance
Perceive current sound
- Encode sound length
Retrieve previous sound
Compare sounds
based on length
Respond
Same / Different
15
Task
|
Study
|
Model
|
Discussion
|
Future
Errors
Perceive current sound
Encode sound length
Encode sound pitch
Retrieve previous sound
Expect matching pitch
Allow partial matching
Compare sounds
based on length
Respond
Same / Different
16
Task
|
Study
|
Model
|
Discussion
|
Future
Deterioration at depth
:ans = 0.1; 0.2; 0.25
Perceive current sound
- Encode sound length
Compare sounds
based on length
Respond
Same / Different
Retrieve previous sound
:mp = 1.0; 1.5; 2.0
Perceive current sound
Encode sound length
Encode sound pitch
Retrieve previous sound
Expect matching pitch
Allow partial matching
Compare sounds
based on length
Respond
Same / Different
17
|
Study
|
Model
|
Discussion
|
Future
Model fit
Performance relative to pre-dive
1
0.95
0.9
0.85
Ratio
Task
0.8
0.75
0.7
0.65
0.6
0.55
0.5
Pre-dive
Dive 190 fsw
Dive 280 fsw
18
|
Study
|
Model
|
Discussion
|
Future
Model predictions
Error rate by pitch
0.6
0.5
0.4
Error rate
Task
0.3
0.2
widely deviant
0.1
moderately deviant
slightly deviant
0
prev.
standard
prev. slightly
deviant
standard
prev.
moderately
deviant
prev. widely
deviant
19
Task
|
Study
|
Model
|
Discussion
|
Future
Discussion

Why 2 parameters and not just one

Only activation noise (ans):


No correlation pitch deviation – accuracy
Only mismatch penalty (mp):

No errors in standard pitch condition

Needs data to check predictions

What does the model tell us?

Need to explain parameter change in mechanistic
terms
20
Task
|
Study
|
Model
|
Discussion
|
Future
Future research

Learning

Affective effects associated with
performance decrement

Anxiety, euphoria, overconfidence, etc.

Transfer and generalization

Brain imaging, brain stimulation
21
Task
|
Study
|
Model
|
Discussion
|
Future
Learning

Learning effects:

Accuracy



Ignore the distracting feature of stimulus
No RT effect

Fast paced

Wait for sounds to finish
Transfer to different tasks
22
Task
|
Study
|
Model
|
Discussion
|
Future
Acknowledgements

Data collection study funded by ONR

Modeling study funded by NSMRL

Thanks to Dan Bothell for changes to
the ACT-R auditory module

Thanks to Dario Salvucci and Frank
Ritter for introducing me to Michael Qin
23



Thank you for your attention!
Questions?
24