Modeling Prior and Retrospective f. Awareness of Actions. WS(c). SS(c). SR(c). PAwr(a i. ,b i. ,c,s) RAwr(a i. ,b i. ,c,s). â. EO(a i. ,b i. ,c,s). PO(a i. ,b i. ,c,s). RO(a.
A Computational Cognitive Model A Computational Cognitive Model for Intentional Inhibition of Actions
Dilhan J. Thilakarathne J T Jan Treur Thilakarathne, D.J. & Treur, J. (2013). A Computational Cognitive Model for Intentional Inhibition of Actions. In C.S. Teh, H.R. Chae, S.A.Z. Adruce P N Anding C J Chen N A Aziz & K W Tan (Eds ) Proceedings Adruce, P.N. Anding, C.J. Chen, N.A. Aziz & K.W. Tan (Eds.), Proceedings of the 9th International Conference on Cognitive Science, ICCS'13. Procedia ‐ Social and Behavioral Sciences (pp. 63‐72).
Motivation
• But what will happen with the age? 2
Motivation
3
Motivation
4
Motivation
http://www.youtube.com/watch?v=dIBBDuQrd‐I 5
Motivation • Both psychology and neuroscience confirm that ot psyc o ogy a d eu osc e ce co t at the brain circuits for inhibiting action are just as prominent and important as those for initiating action • Inhibition is an essential aspect of: – – – – –
behaviour regulation self‐control delayed gratification delayed gratification social contracts trust in others trust in others 6
Motivation • Studies Studies have shown that internally generated have shown that internally generated and an externally triggered route to action (Goldberg 1985; Jahanshahi et al., 1995; (Goldberg, 1985; Jahanshahi et al 1995; Jenkins et al., 2000). – medial frontal system for internally generated medial frontal system for internally generated action, centered on the pre‐SMA (supplementary motor area), and a more lateral parietal‐premotor ), p p system for externally triggered action
7
Motivation • Go / No Go / No‐Go Go decision decision
8
Action inhibition Action inhibition • Inhibition: – Unintentional inhibition • May May occur prior to conscious awareness occur prior to conscious awareness when when irrelevant information is automatically activated in conjunction with relevant information. – Ex. Non directed forgetting effect
– Intentional inhibition •M Must consciously decide t i l d id that the information is th t th i f ti i irrelevant and then inhibit its activation
9
What is intentional inhibition What is intentional inhibition • "intentional intentional inhibition inhibition" refers to the capacity refers to the capacity to voluntarily suspend or inhibit an action – He became drunk and lost his inhibitions He became drunk and lost his inhibitions – I must suppress my urge to eat unhealthy foods – You are writing an email to your boss, and just You are writing an email to your boss and just before you click the “send” button, you seem to hear a voice in your head that says “do hear a voice in your head that says do you really you really want to send that?”, and you hold back
10
What is intentional inhibition What is intentional inhibition • The The interplay between a positive potential interplay between a positive potential selection of an action, and the negative impacts of the same action is addressed impacts of the same action is addressed
11
What is intentional inhibition What is intentional inhibition • In In this work distinction of action formation this work distinction of action formation and intentional inhibition is achieved through: – Performative desires: short term desires: short term – Constitutive desires: long term
• If If an action satisfies a performative ti ti fi f ti desire but d i b t not a constitutive desire, intentional inhibition may prevent the action from becoming t th ti f b i executed 12
Mechanism for Intentional Inhibition Mechanism for Intentional Inhibition neural brake
distal‐loop, which serves to check whether the current action goal itself is or is not appropriate. This additional loop is not simply a loop is not simply a hierarchically prior mechanism that selects what motor loop action to perform next. Rather, it makes an additional set of predictions about distal loop longer term consequences and implications of the and implications of the current action, in addition to predicting whether it will achieve the proximate goal. Fil i h E, Kühn Filevich E Küh S, Haggard P. Intentional inhibition in human action: The power of S H d P I t ti l i hibiti i h ti Th f “no”. Neuroscience & Biobehavioral Reviews 2012; 36(4): 1107‐1118 13
Computational modeling Computational modeling • An An important role both in the execution important role both in the execution decisions for an action, and in its attribution, is played by the prediction and valuation of is played by the prediction and valuation of the (expected) effects of the action, based on internal simulation starting from the internal simulation starting from the preparation of the action (e.g., Wolpert 1997, Haggard 2008) Haggard 2008) – As‐if body loop (Damasio, 1994; 1999)
14
Modeling prediction and valuation Modeling prediction and valuation
parallel action simulation and evaluation: a1, a2, ..., an WS(s)
SS(s)
SR(s) PA(ai) SR(bi) F(bi)
as if body loop Damasio, 1994; 1999; 2010
15
Modeling action execution Modeling action execution predicted effect and the sensed actual effect are not simply compared or matched, but in fact are added to each other in some fact are added to each other in some integration process (e.g., Moore and Haggard, 2008; Synofzik et al., 2010; Voss et al., 2010) WS(s)
SS(s)
EA(ai)
SR(s) PA(ai)
WS(bi)
SS(bi)
SR(bi) F(bi)
body loop 16
Modeling prediction and valuation Modeling prediction and valuation • If If these predicted effects are valued as these predicted effects are valued as satisfactory, this may entail a ‘go’ decision for the execution of the action thus exerting the execution of the action, thus exerting control over action execution. In contrast, less satisfactory predicted effects may lead to a ‘no satisfactory predicted effects may lead to a no go’ decision.
17
Prior and retrospective effects Prior and retrospective effects • Importance Importance of prior and retrospective effects of prior and retrospective effects relative to action execution (Treur 2011) • Awareness
D'ostilio, K., Garraux G.: Brain mechanisms underlying automatic and unconscious control of motor action. Frontiers in Human Neuroscience. 6, (2012) 18
Modeling Prior and Retrospective ownership h WS(c)
SS(c)
SR(c) EO(ai,bi,c,s)
PO(ai,bi,c,s)
WS(s)
SS(s)
―
RO(ai,bi,c,s)
EA(ai)
SR(s) PA(ai)
WS(bi)
SS(bi)
SR(bi) F(bi)
Treur, J. (2011). A Cognitive Agent Model Incorporating Prior and Retrospective Ownership States for Actions. In T. Walsh (Ed.), Proceedings of the Twenty‐Second International Joint Conference on Artificial Intelligence, IJCAI'11 (pp. 1743‐1749).
19
Modeling Prior and Retrospective Awareness of Actions f WS(c)
SS(c)
SR(c)
PAwr(ai,bi,c,s) RAwr(ai,bi,c,s)
―
EO(ai,bi,c,s)
PO(ai,bi,c,s)
WS(s)
SS(s)
―
RO(ai,bi,c,s)
EA(ai)
SR(s) PA(ai)
WS(bi)
SS(bi)
SR(bi) F(bi)
20 Thilakarathne, D.J. & Treur, J. (2013). Modelling Prior and Retrospective Awareness of Actions. In proc. of 5th International Work‐ Conference on the Interplay Between Natural and Artificial Computation, Part I. Vol. 7930. Lecture Notes in Computer Science (pp. 62‐73).
Modeling Prior and Retrospective Awareness of Actions f WS(c)
SS(c)
SR(c)
PAwr(ai,bi,c,s) RAwr(ai,bi,c,s)
―
EO(ai,bi,c,s)
PO(ai,bi,c,s)
WS(s)
SS(s)
―
RO(ai,bi,c,s)
EA(ai)
SR(s) PA(ai)
WS(bi)
SS(bi)
SR(bi) F(bi)
21 Thilakarathne, D.J. & Treur, J. (2013). Modelling Prior and Retrospective Awareness of Actions. In proc. of 5th International Work‐ Conference on the Interplay Between Natural and Artificial Computation, Part I. Vol. 7930. Lecture Notes in Computer Science (pp. 62‐73).
Modeling intentional inhibition of action execution The impact prediction loop involves awareness states while the effect prediction loop mainly demonstrates unconscious behavior Prior Ownership and Retrospective Prior Ownership and Retrospective Ownership states are considered unconscious ownership states and the Prior Awareness and Retrospective Awareness states as conscious states as conscious ownership states Rejecting an action due to less satisfactory valuation is different from intentional inhibition intentional inhibition Unintentional inhibition occurs prior to conscious awareness.
Thilakarathne, D.J. & Treur, J. (2013). A Computational Cognitive Model for Intentional Inhibition of Actions. In C.S. Teh, H.R. Chae, S.A.Z. Adruce, P.N. Anding, C.J. Chen, N.A. Aziz & K.W. Tan (Eds.), Proceedings of the 9th International Conference on Cognitive 22 Science. Procedia ‐ Social and Behavioral Sciences (pp. 63‐72).
Description of the Computational Cognitive Model d l
23
Simulation scenarios Simulation scenarios • Satisfactory Satisfactory predicted action effect but predicted action effect but intentionally inhibited action • Satisfactorily predicted action gets executed Satisfactorily predicted action gets executed • Satisfactorily predicted action get executes while impact prediction is disabled hil i di i i di bl d • Model behavior in total unconscious mode
24
Parameter values of the model Parameter values of the model
25
Selecting values for the connection weights h • Analytically driven approach was used – First a set of scenarios (which should be co‐related) is identified for which at least some (fuzzy level) pattern can be identified. – For a selected scenario the weight values are calibrated to simulate a pattern as expected. pattern as expected. – Once parameter values are obtained for the selected scenario, the same values are used for another scenario (this new scenario should require changes to very few parameters in order to adapt to it). – If the previously identified values provide simulation results for the If th i l id tifi d l id i l ti lt f th new scenario as expected, then the previously obtained parameter values become more justified. – If, in contrast, the simulation results for the new scenario are not as expected, it is required to change some of the selected parameter values until the simulations for this new scenario give results as expected. – Repeat this for all scenarios Repeat this for all scenarios 26
Simulations • Satisfactory Satisfactory predicted action effect but predicted action effect but intentionally inhibited action 12 1.2 1 0.8 0.6 0.4 02 0.2 0 1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
PA(a_obj)
PA(a_sub)
SR(b_obj)
SR(b_sub)
PD(b)
CD(b)
F(b_obj)
F(b_sub)
PO(…)
PAwr(…)
EA(a)
RO(…)
RAwr(…)
EO(…)
81
86
27
Simulations • Satisfactorily predicted action gets executed Satisfactorily predicted action gets executed 1.2
1
0.8
06 0.6
0.4
0.2
0 1
5
9
13
17
21
25
29
33
37
41
45
49
53
57
61
65
69
73
77
PA(a obj) PA(a_obj)
PA(a sub) PA(a_sub)
SR(b obj) SR(b_obj)
SR(b sub) SR(b_sub)
PD(b)
CD(b)
F(b obj) F(b_obj)
F(b_sub)
PO(…)
PAwr(…)
EA(a)
RO(…)
RAwr(…)
EO(…)
81
85
89
28
Simulations • Satisfactorily Satisfactorily predicted action get executes predicted action get executes while impact prediction is disabled* 1 09 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1
5
9
13
17
21
25
29
33
37
41
45
49
53
57
61
65
69
73
77
PA(a_obj)
PA(a_sub)
SR(b_obj)
SR(b_sub)
PD(b)
CD(b)
F(b_obj)
F(b_sub)
PO(…)
PAwr(…)
EA(a)
RO(…)
RAwr(…)
EO(…)
81
85
89
29 *Thilakarathne, D.J. & Treur, J. (2013). Modelling Prior and Retrospective Awareness of Actions. In proc. of 5th International Work‐ Conference on the Interplay Between Natural and Artificial Computation, Part I. Vol. 7930. Lecture Notes in Computer Science (pp. 62‐73).
Simulations • Model behavior in total unconscious mode Model behavior in total unconscious mode
30
Conclusion • The agent model presented here was inspired by g p p y cognitive and neurological evidences, and has shown the combined impact from: – intentional inhibition intentional inhibition – action awareness – action ownership p
• In parallel to the positive action selection process, the intentional inhibition process evaluates the possible negative influence of the current action selection from ti i fl f th t ti l ti f the long term perspective and may lead to abandoning the action 31
Conclusion • The The interplay between conscious and interplay between conscious and unconscious processes achieved through awareness (conscious) and ownership ( ) p (unconscious) states respectively • The experiments have highlighted the fact that if The experiments have highlighted the fact that if intentionally an action was abandoned it takes relatively more time to get settled with the y g original feelings compared to the same when the action got successfully executed 32
Conclusion • The The Anarchic Hand Syndrome (AHS) can be Anarchic Hand Syndrome (AHS) can be simulated analogically by considering Scenario 1 and 2 1 and 2 • Possible application domains: – decision making d ii ki – behavioral management – emotional control l l – simulations for clinical disorders and therapies 33
Thank You Thank You
34