MotivationBased Behaviour

1
Motivation-Based Behaviour

• patterns of behaviour in which the agent
• takes the initiative (without observable
  direct reason)
• does not show expected behaviour (because it
  is not motivated)
• viewed from external and internal viewpoint
• philosophical perspectives
• behaviourist
• functionalist
• instrumentalist
• realist

2
Environmental Complexity Thesis

• stimulus-response behaviour
• works only with visible food
• delayed response behaviour
• food becoming invisible is taken into account
• motivation-based behaviour
• food becoming absent is taken into account
• never visible food is taken into account

3
Experimental Setting

p1
p2
4
Aristotle On the Motion of Animals (1)

• Now we see that the living creature is moved by
  intellect, imagination, purpose, wish, and
  appetite. And all these are reducible to mind and
  desire.
• (De Motu Animalium, 350 BC, Part 6)

Now we see that the living creature is moved by
intellect, imagination, purpose, wish, and
appetite. And all these are reducible to mind and
desire. (De Motu Animalium, 350 BC, Part 6)
5
Aristotle On the Motion of Animals (2)

• But how is it that thought (viz. sense,
  imagination, and thought proper) is sometimes
  followed by action, sometimes not sometimes by
  movement, sometimes not?
• ()
• here the two premisses result in a conclusion
  which is an action - for example, one conceives
  that every man ought to walk, one is a man
  oneself straightaway one walks
• ()
• Now that the action is the conclusion is clear.
  But the premisses of action are of two kinds, of
  the good and of the possible.
• (De Motu Animalium, 350 BC, Part 7)

But how is it that thought (viz. sense,
imagination, and thought proper) is sometimes
followed by action, sometimes not sometimes by
movement, sometimes not? () here the two
premisses result in a conclusion which is an
action - for example, one conceives that every
man ought to walk, one is a man oneself
straightaway one walks () Now that the action
is the conclusion is clear. But the premisses of
action are of two kinds, of the good and of the
possible. (De Motu Animalium, 350 BC, Part 7)
6
Aristotle On the Motion of Animals (3)

• And as in some cases of speculative inquiry we
  suppress one premise so here the mind does not
  stop to consider at all an obvious minor premise
  for example if walking is good for man, one does
  not dwell upon the minor 'I am a man'.
• And so what we do without reflection, we do
  quickly.
• For when a man actualizes himself in relation to
  his object either by perceiving, or imagining or
  conceiving it, what he desires he does at once.
• (De Motu Animalium, 350 BC, Part 7)

And as in some cases of speculative inquiry we
suppress one premise so here the mind does not
stop to consider at all an obvious minor premise
for example if walking is good for man, one does
not dwell upon the minor 'I am a man'. And so
what we do without reflection, we do quickly.
For when a man actualizes himself in relation to
his object either by perceiving, or imagining or
conceiving it, what he desires he does at once.
(De Motu Animalium, 350 BC, Part 7)
7
Aristotle On the Motion of Animals (4)

• I want to drink, says appetite this is drink,
  says sense or imagination or mind straightway I
  drink.
• In this way living creatures are impelled to move
  and to act, and desire is the last or immediate
  cause of movement, and desire arises after
  perception or after imagination and conception.
• And things that desire to act now create and now
  act under the influence of appetite or impulse or
  of desire or wish.
• (De Motu Animalium, 350 BC, Part 7)

I want to drink, says appetite this is drink,
says sense or imagination or mind straightway I
drink. In this way living creatures are impelled
to move and to act, and desire is the last or
immediate cause of movement, and desire arises
after perception or after imagination and
conception. And things that desire to act now
create and now act under the influence of
appetite or impulse or of desire or wish. (De
Motu Animalium, 350 BC, Part 7)
8
The Pro-Active Animal C

p1
p2
9
Comparing Agent Models A, B and C

C2
10
Questions

• What is the pattern in this observed behaviour ?
• Which external dynamic properties can be
  expressed that characterize the pattern behind
  these traces ?
• Which assumed internal state properties generate
  this externally observed behaviour ?
• What is the pattern of dynamics of these internal
  state properties ?
• How can these internal dynamics be characterized
  by dynamic properties ?

11
Dynamic Property
external / internal ? Quality of explanation?

• EMB1
• at any time point t1
• if the agent observes the absence of the screen,
• and time points s1, s2, and s3 exist with
  s1lt s2 lt s3 ? t1 such that
• at s3 the agent observed the presence of food at
  p2, and at s2 the agent observed the absence
  of food at p2, and at s1 the agent observed
  the presence of food at p2,
• then after time point t1 the agent will go to p2
• (the agent starts wanting food if it cannot get
  it
• if wanting food, it reacts immediately if food is
  present)

12
Behaviourist Explanation of Motivation-Based
Behaviour (1)

• Why does the animal go to p2 ?
• The animal goes to p2, because it just observed
  the absence of the screen, and at an earlier
  point in time it observed food at p2, whereas
  at a still earlier point in time it observed
  the absence of food at p2, and at yet another
  still earlier point in time it observed the
  presence of food at p2.

13
Behaviourist Explanation of Motivation-Based
Behaviour (2)

• more a historical type of explanation than based
  on a direct cause (see also the interactivist
  perspective later)
• doubtful whether this is a satisfactory
  behaviourist explanation
• for an explanation based on direct causes
  internal state properties seem to be required

14
Internal State Properties (1)

• motivational attitudes
• desires
• intentions
• informational attitudes (e.g., beliefs)
• additional reasons
• opportunities

15
Internal State Properties (2)
agent desires A
history of the agent
w.r.t. interaction with
agent intends A
external world
agent performs A
agent believes it has an
additional reason to pursue A
agent believes it has
an opportunity to do A
history of the agent
w.r.t. obtained information
on world state
16
Internal Dynamics
d
i
r
o
17
Internal Dynamic Properties for Beliefs

• IDR1 (Belief Generation)
• for all time points
• if the agent observes that food is present at p2,
• then internal state property b1 will hold
• IDR4 (Belief Persistence)
• for all time points t1 and t2 with t1ltt2
• if internal state property b1 holds at t1,
• and between t1 and t2 the agent does not
  observe that food is not present at p2,
• then internal state property b1 holds at t2

18
Internal Dynamic Properties for Action and
Intention Generation

• IMB2 (Intention Generation)
• at any point in time
• if the internal state property d holds,
• and the internal state property b1 holds
• then the internal state property i will hold
• IMB3 (Action Generation)
• at any point in time
• if the internal state property i holds,
• and the agent observes the absence of the screen
  
• then the agent will go to p2

19
Internal Dynamic Property for Desire

• IMB1 (Desire Generation)
• at any point in time
• if the agent observes the absence of food at p2,
• and an earlier time point exists at which
• the agent observed the presence of food at p2,
• then the internal state property d will hold
• Notice this is not an executable property.

20
Persistence Properties

• for all time points
• if internal state property X holds,
• then for every later time point X holds
• IMB4 Desire Persistence X d
• IMB5 Intention Persistence X i
• IMB6 Predesire Persistence X p

21
Conditional Persistence Property

• IMB4
• for all time points t1 and t2
• if internal state property d holds at t1,
  and between t1 and t2 no eat action occurred
• then internal state property d holds at t2

22
Internal Dynamics Graphical Form
23
Animal Behaviour Internal Traces
24
Internal Dynamics Replacing IMB1 by Executable
Properties

• Add internal state property p (predesire)
  dynamics
• IMB1a
• at any point in time
• if the agent observes the presence of food at p2,
• then the internal state property p will hold
• IMB1b
• at any point in time
• if the agent observes the absence of food at p2,
• and the internal state property p holds,
• then the internal state property d will hold

25
Animal Behaviour Internal Traces
26
Animal Behaviour Internal Traces
27
Pain Example

• pain tissue damage detector
• During evolution, nature has developed different
  mechanisms to realize this function.

28
Pain Example Mediating Role

• input tissue damage
• leads to
• pain leads to
• output ouch!
• moving

29
Functional Role Intention
30
Functional Role Intention
31
Functional Role Desire
32
Functional Role Desire
33
Functional Role Belief
34
Functional Role Pre-Desire
35
Functional Roles

• no independent specification of dynamics of one
  internal state property related to input and
  output dynamics
• the functional roles of the internal state
  properties as specified by the dynamic properties
  define their dynamics as a whole (holistic view)

36
Logical Relationships between Internal and
External Dynamic Properties

• Predesire Generation Desire Generation
  Intention Generation
• Action Generation Belief Generation
• Predesire Persistence Desire Persistence
  
• Intention Persistence Belief Persistence
• ? Motivation-Based Behaviour
• IMB1ab IMB2
• IMB3 IDR1
• IDR4 IMB4
• IMB5 IMB6
• ? EMB1

37
(Iterated) Explanation of Motivation-Based
Behaviour (1)

• Why does the animal go to p2 ?
• The animal goes to p2, because it just observed
  the absence of the screen, and it had the
• intention to go to p2.
• Why did the animal have this intention to go to
  p2 ?
• The animal had the intention to go to p2,
  because it had the desire for food and the
• belief that there was food at p2.

38
(Iterated) Explanation of Motivation-Based
Behaviour (2)

• Why did the animal have the desire for food and
  why did it believe that food was present at p2 ?
• The animal desired food, because it had this
• desire for a longer time already. The same
  holds for the belief that food was
• present at p2.

39
(Iterated) Explanation of Motivation-Based
Behaviour (3)

• But then, why did these desire and belief start
  to be there ?
• The desire started to be there because the
• animal observed that there was no food at p2,
  and the animal had a predesire.
• The belief that food is present at p2 started to
  be there, because the animal observed the
• presence of food at p2.

40
(Iterated) Explanation of Motivation-Based
Behaviour (4)

• Why did the animal have this predesire ?
• The animal had this predesire, because it had
• this predesire persisting for a longer time
• already.
• Why did this predesire start to be there ?
• The predesire started to be there because the
  animal observed the presence of food at p2.

41
Motivation-Based Behaviour Summary

• externally observable motivation-based behaviour
  characterised by dynamic properties
• characterising desires may require nontrivial
  references to past input and/or output
• dynamics of appropriate internal state properties
  identified by (executable) dynamic properties

42
Motivation-Based Behaviour Summary

• relationships between internal dynamic properties
  and external dynamic properties
• enable to explain, predict or simulate
  motivation-based behaviour based on (executable)
  internal dynamic properties e.g.,
• the agent does this, because it desires and
  it believes that
• the agent desires , because

43
Intentional Stance

• An example of an instrumentalist view is
  Dennetts intentional stance
• attribution of intentional state properties to
  other agents
• intentional strategy to explain and predict
  behaviour (Dennett, 1987)

44
Physical Stance

• This can be opposed to the physical stance
• (realist view)
• determine the physical constitution of a system
• use knowledge of laws of physics to predict the
  behaviour depending on the inputs
• often useful for engineers

45
Intentional Stance

• treat the system as a rational agent
• figure out what beliefs it has, given its place
  in the world and its purpose
• figure out what desires it ought to have, given
  its place in the world and its purpose
• predict that the agent will behave according to
  its desires and beliefs
• validity of the intentional stance

46
Examples

• a better trap for an animal can be designed if it
  is taken into account what it wants and what it
  believes
• a chess programme will not take your knight
  because it believes that then you will take its
  rook and it wants to keep its rook
• a thermostat will turn off the heater if it
  believes that the room has reached the desired
  temperature
• more (human) examples will follow

47
Advantages

• An intentional stance description abstracts from
  specific aspects of physical realisation
  therefore
• the description is more concise and transparent
• it is more generic it applies to more cases, in
  different physical realisations

48
Why Does it Work ?

• patterns of behaviour of an observed agent form
  an empirical basis
• temporal dependencies between the intentional
  notions and these observable behavioural 'real
  world' patterns, and between the intentional
  notions themselves are crucial.

49
Example Animal Behaviour

• an observed (animal) agent receives observation
  input on - the availability of food - the
  presence or absence of a screen
• depending on the circumstances it can decide to
  eat the food

50
Motivation-Based BehaviourExample Traces
 
 
51
Interpretation Criteria Intention

• For no screen as opportunity
• The agent has the intention to eat at each time
  point that
• food is visible, and
• in the past the agent experienced that visible
  food can suddenly disappear.

52
Interpretation Criteria Desire
For food present as additional reason for the
intention The agent has the desire to eat at
each time point that in the past the agent
experienced that visible food can suddenly
disappear.
53
Examples within an Organisation

• How to influence in anticipation which intentions
  will or will not arise in another participant
  within an organisation.
• Prediction based on the intentional stance,
  followed by an intervention to avoid the
  prediction to become true.

54
Context

• coordinated dynamics by specifications of
  constraints on role behaviour and on interaction
  dynamics
• these specifications usually do not completely
  prescribe behaviours
• some space of freedom in behaviour and personal
  initiative is allowed
• considered from the instrumentalist perspective

55
Examples of these Phenomena in Human Organisations

• Firing intention
• An employee has done something very important
  very wrong, and deliberates whether or not to
  tell his manager
• 'If he believes that I am the cause of the
  problems, he will try to fire me.'

56
Innovation Intention

• An employee has encountered a recurring problem
  within the organisation, and knows a solution for
  this problem, on which he would like to work. He
  deliberates about how to propose this to his
  manager.

57
Deliberation of the Employee

• 'If I tell this solution immediately he will not
  believe that the problem is worth working on it.
  If I make him aware of the problem, and do not
  tell a solution, he only will start to think
  himself about it for a while, without finding a
  solution, and then forget about it. If I make
  him aware of the problem and give some hints that
  direct him to a (my) solution, he will believe he
  contributed to a solution himself and want me to
  work (further) on it.'

58
Task Avoidance Intention

• A manager observes that a specific employee often
  shows avoidance behaviour in cases that a task is
  only asked shortly before its deadline.
• The manager deliberates
• 'If I know beforehand the possibility that a
  last-minute task will occur, I can tell him the
  possibility in advance, and in addition point out
  that I need his unique expertise for the task, in
  order to avoid the behaviour that he tries to
  avoid the task when it actually comes up.'

59
Instrumentalist Analysis of the Task Avoidance
Case

• At some time point the employee has the desire to
  reject the task if
• the employee just heard the request to perform
  the task
• the employee just heard that the task has to be
  finished soon
• the employee did not hear of the possibility of
  the task at any earlier time point

60
Intention Attribution

• At some time point the employee has the intention
  to reject the task if
• the employee has the desire to avoid the task
• the employee believes that colleagues are capable
  of doing the task
• the employee believes that colleagues are not
  full-booked

61
Action

• At some time point the employee performs the
  action to reject the task if
• the employee has the intention to avoid the task
• the employee believes that the employee's own
  agenda is full-booked

62
How to Avoid Avoidance Behaviour

• Avoid one of the conditions listed above
• Avoiding the desire to occur
• communicate in advance to the employee that
  possibly a last minute task will occur

63
Avoiding the Intention to Occur (Given that the
Desire Occurs)

• Tell the employee that he is the only one with
  the required expertise
• Tell the employee that colleagues are full-booked

64
Avoiding the Action to Occur (Given that the
Intention Occurs)

• Take one of the (perhaps less interesting) tasks
  from his agenda and re-allocate it to a colleague

65
Motivation Theories for Organizations
Adapted from T.R.Mitchell and D.Daniels (2002)
66
Characteristics of agents (1)
The following division of needs is identified in
social science (1) extrinsic needs associated
with biological comfort and material
rewards (2) social interaction needs that
refer to the desire for social
approval, affiliation and companionship (3)
intrinsic needs that concern the desires for
self- development, self-actualization,
mastery and challenge.
67
Characteristics of agents (2)
Goals associated with intrinsic needs are often
satisfied by intrinsic rewards that are a natural
consequence of the agent behavior related to the
execution of a task Externally provided rewards
(e.g., salary, bonuses, group acceptance) serve
to the satisfaction of goals related to extrinsic
and social interaction needs
68
Modeling the motivation of an agent (1)

• For computational modeling of motivation the
  expectancy theory by Vroom can be adopted
• According to the expectancy theory by Vroom, when
  an individual evaluates alternative possibilities
  to act, s/he explicitly or implicitly makes
  estimations for the following factors
  expectancy, instrumentality, and valence

69
Modeling the motivation of an agent (2)
Expectancy (E) refers to the individual's belief
about the likelihood that a particular act will
be followed by a particular outcome Expectancy
of an agent related to successful task execution
is determined by the characteristics of the task
and the agent, and by the organizational and
environmental conditions
70
Modeling the motivation of an agent (3)
Instrumentality (I) is a belief concerning the
likelihood of a first level outcome resulting
into a particular second level outcome its value
varies between -1 and 1. A second level
outcome represents a desired (or avoided) by an
agent state of affairs that is reflected in an
agent's goal(s)
71
Modeling the motivation of an agent (4)
The agents experience gained by the execution
of tasks influences the values of expectancies
and instrumentalities associated with these
tasks. Valence (V) refers to the strength of
the individual's desire for an outcome or state
of affairs.
72
Modeling the motivation of an agent (5)
In the Vroom model the force on an individual to
perform an act is
73
Example
74
Conclusion

• agents within an organisation try to affect, in
  an anticipatory manner, the circumstances that
  may lead to the generation of other agent's
  beliefs, desires and intentions.
• the capability of performing such anticipatory
  reasoning is crucial for functioning in human
  society