a first step Hawk-Dove-Assessor Game

Chapter 21: Conflict
Resolution
Single threat display
Why so many threat displays?
BGYD45 2003:8
1
Hawk-Dove Game: a first step
Hawk
Hawk ½(V – D)
Dove
zero
½V
2
Hawk-Dove-Assessor Game
Dove
V
BGYD45 2003:8
• New strategies:
– Bully
If V > D, then it always
pays to be a hawk: pure
ESS
• Escalate on first move, continue if opponent
displays, retreat if they escalate
– Retaliator
But if V < D, then no
single best strategy:
mixed ESS
• Display first, escalate if opponent does
– Prober-retaliator
• Like retaliator, but occasionally throws in an
escalation
A bit too simple, but does predict that high costs favour
ritualization and escalation more likely with low costs.
BGYD45 2003:8
3
BGYD45 2003:8
4
1
Hawk-Dove-Assessor Game
Assessor Strategy
• New strategies:
• Assessors display first, evaluate whether
they would win or lose from information
provided by the display, and then retreat if
they assess themselves to be the loser.
– Bully
• Bluffing – doesn’t do well
– Retaliator
• Successful
– Prober-retaliator
• Successful
• Responsiveness to opponent is better than
all-or-none strategies.
– Better still, if the responsiveness includes an
estimate of chances of winning: Assessment
BGYD45 2003:8
5
Correlated asymmetry
– Assessor strategy: if larger play hawk, if smaller play
dove
– If owner and intruder are equally frequent and get equal
payoffs:
BGYD45 2003:8
Dove
V
V/2
3V/4
6
Assessor Threat Displays
• Opponents differ in RHP
• Example: hawk - dove - assessor
Opponent:
Hawk
Assessor
(V- D)/2
Actor:
Hawk
Dove
0
Assessor V/2
BGYD45 2003:8
• Costless and accurate don’t usually go
together.
• What is effect of inaccuracy
– Or: how much accuracy is necessary?
• Not that much.
(V- D)/2
V/4
V/2
When there is a cost to fighting, Assessor is pure ESS
assuming that assessment is costless and accurate
7
– Enough to be just better than random
• If there is a cheap unbluffable display that
provides reasonably accurate information
on, for e.g., body size, this is all you need.
BGYD45 2003:8
8
2
Assessor
• Over a wide range of reasonable values
for fighting costs and asymmetries,
Assessor is a pure ESS over Hawk and
Dove. D can be > or < V.
• A single threat behaviour will only be
stable when the display is constrained to
reveal the truth about an asymmetry that
determines the probability of winning.
Usually index signals.
BGYD45 2003:8
9
Calling in toads
BGYD45 2003:8
10
Variable Length Contests
• One display repeated a variable number of
times.
• Symmetric and asymmetric War of Attrition
BGYD45 2003:8
11
BGYD45 2003:8
12
3
War of attrition - the waiting
game
War of attrition - Payoff matrix
• Assumptions
– Resource cannot be shared
– Cost of display increases with length of contest
– No information is received during contest and
opponents are symmetrical
– Winner is the contestant willing to accept the higher
cost
– The cost to both contestants equals the cost
acceptable to the loser
– The range of actions of each contestant is
continuous
BGYD45 2003:8
13
Probabilistic strategy
Play all values of x
with varying
probability – lower
values more
probable.
P(x)
BGYD45 2003:8
Payoff to:
Actor :
xA > xB
Opponent: x A < x B
Player A
V - kx B
- kx A
Player B
- kx B
V - kx A
No pure ESS is possible, since an opponent that displayed a little
bit longer would have higher fitness
Solution is a mixed ESS where the probability of leaving at any
time is a constant. The times an individual stays should be
distributed as a negative exponential.
BGYD45 2003:8
14
Asymmetric war-of-attrition
P(x) = k/V*exp( -kx/ V )
k/V
xi = amount of time individual i displays
k = rate at which costs are expended
V = value of resource
15
• If animals experience different costs of
display or the resource differs in value to
them, the game is asymmetric
• Which player has the largest V/k will win,
but this may not be known
• This may lead to two different giving-uptime strategies
BGYD45 2003:8
16
4
War-of-attrition solutions
War of Attrition with Assessment
• Initial display is how contestants decide their
role, then decide on their “bid” accordingly
• Contest costly for both, so point at which V=k
will determine maximum investment
• Small asymmetry, greater uncertainty of
assessment, increased average and variance of
contest duration
• Greater resource value, greater contest duration
BGYD45 2003:8
17
BGYD45 2003:8
Fight duration and resource value in
newts
Signals and assessment of roles
Males fight longer over
larger females. Larger
females carry more eggs
and are, therefore, more
valuable.
• Assessment must be imperfect
• Displays should not reveal intention or
motivation (only potential strength)
BGYD45 2003:8
18
19
BGYD45 2003:8
20
5
Bowl & doily spiders
Bowl & doily spider
Resource value equal
BGYD45 2003:8
21
Sequential assessment
BGYD45 2003:8
22
Sequential Assessment
• Fixed “bid” aspect of WoA not very realistic
xna = 1/n Σ ?AB + zia
– What about updating estimate of chances as contest
continues?
Random error
• Assumptions
– Animals display in order to acquire information about each
other’s fighting ability and resource value
– Fights only occur when animals are closely matched
?AB = - ? BA relative fighting ability
• Predictions
Each repeated display gives one “sample” of
asymmetry. Accuracy increases with number of
samples: SDx = s /sqrt(n )
– Fight duration increases as the asymmetry in fighting ability
decreases and/or as resource value increases
– The cost of a fight increases as the asymmetry in fighting
ability decreases
– Probability of winning increases with asymmetry
BGYD45 2003:8
23
BGYD45 2003:8
24
6
0.004
?±z
0.000
0.002
dnorm(1:500, 250, 50)
0.006
0.008
Sequential assessment ESS
0
BGYD45 2003:8
25
BGYD45 2003:8
100
200
300
400
500
Index
26
Ownership effects in spider fights
Sequential assessment
• Predictions similar to WoA
– Longer, more variable contest with smaller
asymmetry
– Short contest with large asymmetry
• But more plausible than WoA
– No “sealed bid”
– Outcome determined by relative fighting abilities
– Contestant control their investment as contest
proceeds
BGYD45 2003:8
27
BGYD45 2003:8
28
7
Ownership effects in spider fights
Fixed sequence contests
• Sequential assessment game with several
different behaviours
– Behaviours occur in a consistent sequence
– Each is an escalation over previous
BGYD45 2003:8
29
30
Assessment in red deer
Jumping spider contest stages
BGYD45 2003:8
BGYD45 2003:8
31
BGYD45 2003:8
32
8
Sequential assessment in
phases
A brief digression
Low frequency: honest, but still maximizing their potential
BGYD45 2003:8
33
Variable sequence contests
34
Why give multiple signals?
• Many different displays, no consistent order
• Each display serves a different function
• Examples: the web-building spider Agelenopsis
with 33 different behaviors used during contests,
penguins with 22 behaviours, and crickets with
12 different behaviours.
• Most behaviours are unilateral
• Behaviours not typically repeated (in a row) –
rather a chain of actions/responses
• Rapid escalation and de- escalation
BGYD45 2003:8
BGYD45 2003:8
– Maybe in crickets
• Signal erosion: threat displays lose effectiveness
as the frequency of bluff increases
– Maybe some theoretical support
• Displays transmit graded information about
display intentions
– Multiple signals may lower potential risk associated
with escalation
– Maybe little blue penguins
35
BGYD45 2003:8
36
9
Little blue penguin
Cricket display costs
Different functions
BGYD45 2003:8
37
BGYD45 2003:8
38
Little blue penguin display repertoire
Little blue penguin fights
• Burrow dwellers
– Similar to crickets (in signalling system):
intermediate repetoire size; little redundancy
within functional categories
• Cave dwellers
– Larger repetoire ; multiple signals within each
functional category
Some populations nest in caves, some in burrows.
• Why?
Cave dwellers use more displays, but have fewer escalated fights.
BGYD45 2003:8
39
BGYD45 2003:8
40
10
More or fewer displays?
More or fewer displays?
• Remember asymmetric WoA
• More displays can be seen as dividing the
signal axis more and more finely
– It pays to use a conventional settlement
• If there is a cheap unbluffable display that
provides reasonably accurate information on, for
e.g., body size, this is all you need.
– No need for complex repetoire with many displays
• But also remember that not all contests are so
simply decided.
– More accurate representation of your relative
position
• The more signals, the smaller the
probability of escalated fight.
– e.g. ownership, resource value differences, etc.
contribute to more contingency in outcome
BGYD45 2003:8
41
BGYD45 2003:8
Penguins again
What’s the point
• Cave dwellers are closer together, with
fewer intervening barriers. Burrow dwellers
more isolated.
• Frequency of conflict three times higher in
cave dwellers, but frequency of escalated
fights is lower.
– Due to more subtle and elaborate signalling
repetoire.
BGYD45 2003:8
42
43
• Diversity of complex behaviours
associated with aggressive contests
– What are the structures of stable resolution
mechanisms?
– How do they arise
• Must find appropriate model to understand
behavioural dynamics.
BGYD45 2003:8
44
11