Real democracy: how honey bees select a new home

1
Real democracy how honey bees select a new home

• Thomas D. Seeley
• Dept. of Neurobiology and Behavior
• Cornell University

2
Democracy --government in which the supreme
power is vested in the people and exercised by
them directly, or indirectly through a system of
representation
3
Democracy --government in which the supreme
power is vested in the people and exercised by
them directly or indirectly through a system of
representation

• Real Democracy
• --when all citizens are legislators, i.e. they
  meet in a face-to-face assembly and bind
  themselves under decisions they make themselves

4

• Athens, Greece 508 BC
• Athens, Vermont (and 1000 other towns in N.E.)
• snowmobiler associations, university
  departments, poker clubs
• honey bee colonies (for 20 million
  years)

5
Democracy can blunder

• No one pretends that democracy is perfect or
  all-wise. Indeed, it has been said that
  democracy is the worst form of Government except
  all those other forms that have been tried from
  time to time.
• --Winston Churchill, 1947
• Speech in the House of Commons

6
Good decision making by democratic groups is not
automatic

• The mass never comes up to the standard of its
  best member, but on the contrary degrades itself
  to a level with the lowest.
• Henry David Thoreau, Journal, 14 March 1838
• Madness is the exception in individuals but the
  rule in groups.
• Friedrich Nietzsche, Beyond Good and Evil, 1886

7
THE WISDOM OF SWARMS How the Many Are Smarter
than The Few
With the right organization, democratic groups
are remarkably intelligent, often smarter than
the smartest individuals in them.
8
A Swarm of Bees

• Reproductive propagule
• One queen bee
• 10,000 worker bees
• 2-3 are active
• (200-300 scout bees)
• 97-98 are quiescent

9

• Citizens? registered voters (100-2500)
• Participants? Town meeting attendees (50-300)
• Agenda items? 10-40
• Voting? raising hand, standing up, secret
  balloting

• Citizens member workers
  (2000-12,000)
• Participants Nest-site scouts (200-300)
• Agenda items? Only 1
• Voting? spending time at a potential home site
  (voting with their wings)

10
Agenda item 1 Where will we build our new home?
11
Pioneering discovery by Martin Lindauerscout
bees report potential home sites with waggle
dances (1955)
Martin Lindauer
Karl von Frisch
12
(No Transcript)
13
1. Angle of waggle run indicates direction.
Coding location information in waggle dance
2. Duration of waggle run indicates distance.
14
Lindauerskey findings

• Initially, bees perform dances for multiple sites
• Before swarm flies away, all dances indicate one
  site
• The swarm flies to the consensus site, moves in
• Therefore, dances on swarm indicate nest sites
• Scouts are holding a kind of plebiscite on the
  swarms new home.

Lindauer (1955) Z. vergl. Physiol. 37263-324.
15
The real estate preferences of bees

• (gt means is preferred to)
• Entrance height 5 gt 1 m
• Entrance area 15 gt 75 sq cm
• Entrance direction south gt north
• Entrance position bottom gt top
• Cavity volume 40 gt 10 liters
• Combs with gt without

16
Does a swarm choose the best of the various
sites that it examines?
Shoals Marine Laboratory Appledore Island, Maine
17
Swarm clustered on mount
18
Variable quality nest site
19
Results (note winner takes all)
Scouts visible at nest box
Time of day
20
How exactly do the scout bees conduct their group
decision making?

• How does democracy work in a honey bee swarm?

21
Detailed eavesdropping on the scout bees
debate on a swarm
22
Videorecord and transcribe waggle dances
23
One 16-hour debate 11 sites, 149 scouts
24

What is the decision evidence? Where is it
accumulating?

• Dances Scout bees
• on swarm? at sites?

Decision-making processconsensus building or
quorum building?
25
Testing the hypothesis of consensus building
Critical prediction Consensus is needed to reach
a decision.
26
Conclusion A consensus among the dancing bees
is not needed to reach a decision.
Testing the hypothesis of consensus sensing
Critical prediction Consensus is needed to reach
a decision.
Swarms took off with dancers split between two
sites!
27

• Decision evidence number of scouts at each site
• Making a decision accumulating a quorum of bees
  (15-20 outside, simultaneously) at one site
• Quorum sensing remains a mystery

28
Swarm must warm up to be able to fly to new
homewarming takes at least 30-60 minutes
Why quorum building, not consensus biulding?
Seeley, Kleinhenz, Bujok Tautz (2003)
Naturwissenschaften 90256-260.
Warm up starts as soon as enough scouts (not all
scouts) have approved of a site boost speed,
maintain accuracy
29
What are the behaviors of the individual scout
bees that underlie the rapid buildup of scouts at
superior sites, and the eventual decline of
scouts at the inferior ones?
30
Tuning of dance duration as a function of site
value
31
A scout makes multiple visits to her site, but
dances less and less strongly after each visit
Blue
Green
Yellow
Orange
Red
32
Decay function for the dances of scout bees
6 5 4 3 2 1 0

Remaining returns to swarm with dancing
33
Why the scout buildup is stronger at a better site
Superb site
907560453015 315 waggle runs
So-so site
3015 45 waggle runs
6 5 4 3 2 1 0
Remaining returns to swarm with dancing
34
Why the scout buildup is stronger at a better site

• Scouts for the better site have a higher per
  capita recruitment (birth) rate and a lower per
  capita abandonment (death) rate.
• Population of scouts for the best site grows most
  rapidly, and ultimately overwhelms, all
  populations for other sites.

Superb site
907560453015 315 waggle runs
So-so site
3015 45 waggle runs
6 5 4 3 2 1 0
Remaining returns to swarm with dancing
35
Dynamics on swarm cluster and at nest sites that
underlie swarm decision making
36
Swarm Smarts(tips for good group decision making)

• Consider a broad array of alternative
• courses of action.
• scouts search widely and report freely
• Avoid the tendency to find rapid agreement.
• scouts conduct a lively competition among
  options
• 3. Promote independence in judging the
  alternatives.
• scouts assess sites independently (no
  imitation/fads)
• 4. Aggregate opinions with both speed and
  accuracy.
• scouts use quorum building (not consensus
  building)

37
(No Transcript)
38

• Athens, Greece 508 BC
• Athens, Vermont (and 1000 other towns in N.E.)
• snowmobile associations, university departments,
  poker clubs
• honey bee colonies
• Kendal at Ithaca?

39

• Collaborators
• Madeleine Beekman (Sydney)
• Brigitte Bujok (Würzburg)
• Barrett Klein (UTexas - Austin)
• Marco Kleinhenz (Würzburg)
• Kevin Passino (Ohio State)
• Jürgen Tautz (Würzburg)
• Kirk Visscher (UC-Riverside)

• Field Assistants
• Susannah Buhrman
• Siobhan Cully
• Robert Fathke
• Benjamin Land
• Kristin Pastor
• Adrian Reich
• Clare Rittschof
• Ethan Wolfson-Seeley
• Ariel Zimmerman

Funding NSF, USDA, National Geographic
Society, Alexander von Humboldt Foundation
Inspiration Prof. Dr. Martin Lindauer (Würzburg)
40
Decision-making processconsensus building or
quorum sensing?
What is the decision evidence? Where is it
accumulating?

• Dancer consensus Scout quorum
• at swarm? at site?

41
Laboratory for experiments withhouse hunting
beesAppledore Island, Maine
Shoals Marine Laboratory Cornell University
University of New Hampshire
42
Testing the hypothesis of quorum sensing

• Critical prediction
• Delaying quorum formation at the chosen site,
  while leaving the rest of the decision-making
  process undisturbed, should delay the reaching of
  a decision.

43
Experimental methods
Each swarm conducted its decision-making process
twice, once with 1 nest box, and once with 5 nest
boxes (or vice-versa).
44
When quorum is reached, scout bees produce an
acoustical signal (worker piping)to stimulate
non-scouts to warm up for flight
Seeley Tautz (2001) J. Comp. Physiol. A 18766
7-676.
45
(No Transcript)
46
Piping starts long before liftoff
Seeley Tautz (2001) J. Comp. Physiol. A
187667-676.
What triggers piping? Quorum of scouts at a site.
47
1 nest-box trials vs. 5 nest-box trials
1400
1500
1600
700 800 900
1000 1100
Seeley Visscher (2004) Behav. Ecol. Sociobiol.
56594-601.

• Slower buildup of scouts at each nest box
• No decrease in dancing at swarm
• Marked delay in time to decision!
• (on average, 3.3 vs 7.4 hours, P lt 0.005)

48
Conclusions

• Decision evidence number of scouts at each site
• Making a decision accumulating a threshold
  number (quorum) of bees at a site
• How bees sense the quorum remains a mystery

49
Decision making by accumulation of evidence
Monkey brain
Bee swarm
Selected nest box
quorum
Nonselected nest box
50
Swarm Smarts!

• Promote diversity of knowledge within
• the group
• scouts search autonomously and report freely
• Avoid tendency to conformity, rapid agreement
• scouts conduct an open competition among
  opinions
• scouts vote independently (no imitation/fads)
• Aggregate opinions with both speed and accuracy
• scouts use quorum sensing,with moderate quorums

51
Leslie matrix model by Mary Myerscough

• Populations of dancers for different sites
• Within each population
• Dancers age
• Dancers reproduce (recruit a new dancer)
• Dancers die (stop dancing)

52
Neurobiology of simple decisions
53
Absolute stopping-rule models
54
The question of social choice How can a group
use the knowledge and opinions possessed by its
members to produce an optimal choice of action
for the group as a whole?
55

• Stimuli

Nest sites
Sensory signals
Waggle dances
Decision making process
Evidence race
Piping signals
Motor commands
Action
Swarm flight
56
Mechanisms of swarm decision making

• scout bees function as sensory units
• waggle dances are sensory signals
• decision evidence is scouts at nest sites (not
  dances at swarm)
• decision evidence accumulates separately for each
  option accumulation is leaky mutual
  inhibition
• decision occurs when evidence total for one
  option exceeds a threshold

57
How do dancers stop dancing for rejected
sites?Do they quit dancing for one site when
they receive news of a better site?
58
Winsors equations for the competitive exclusion
principle dN1/dt r1N1U - a1N1 dN2/dt
r2N2U - a2N2 Ni is the number of scouts
committed to site i U is the number of
uncommitted scouts ri is the rate of recruitment
per scout committed to site i ai is the rate of
abandonment per scout committed to site i
N1r2 / N2r1 e (r1a2 - r2a1)t
59
Decision process is a race between competing
evidence totals
60
House hunting by honey beesa study in group
decision making

• Thomas D. Seeley
• Department of Neurobiology and Behavior
• Cornell University

61
Mechanisms of swarm decision making

• scout bees function as sensory units
• waggle dances are sensory signals
• decision evidence is scouts at nest sites (not
  dances at swarm)
• decision evidence accumulates separately for each
  option accumulation is leaky mutual
  inhibition
• decision occurs when evidence total for one
  option exceeds a threshold

62
Scout bees stimulate non-scouts to warm up with
an acoustical signal wings-together worker
piping
63
Beekeepers quickly hive swarms (freebies)
so home finding by honey bees was long a
deep mystery.
64
Piping/warming takes 30-60 minutes
Seeley, Kleinhenz, Bujok Tautz (2003)
Naturwissenschaften 90256-260.
Why quorum sensing, not consensus sensing? Warm
up starts as soon as enough scouts (not all
scouts) have approved of a site boost speed,
maintain accuracy
65
Friendly competition among coalitions of
committed scouts for the uncommitted scouts
So-so site
Superb site
Site 1 bees
Uncommitted Scout bees
Site 2 bees
r1
r2
a1
a2
N1
N2
U
For each site i dNi/dt NiriU - Niai Bees
need this r1 gt r2 and a1 lt a2
66
Modeling and analysis of swarm cognition
Passino Seeley (2007) Behav. Ecol Sociobiol.
In press
Individual-oriented model, 100 scout bees, 6 sites
67
Notations for model
Nest-site landscape for model
68
Make pseudomutant swarms with modelExample
effects of quorum size
Eq quorum size