2. Swarming
• Describes self‐organizing behavior in popula$ons by
which local interac0ons between decentralized simple
agents can create complex global swarming behavior.
• Every agent is only responsible for its individual
ac$ons.
• Swarm intelligence refers to systems which accomplish
complex global tasks through the simple local
interac$ons of autonomous agents.
• Swarm intelligence relies upon the emergent
proper$es of its components to manifest itself.
Emergence is the process by which complex paAerns
form out of the interac$on of simpler rules.
4. Interrela$ons in swarm
• An individual must maintain an iden$ty
throughout its existence.
• An individual must sa$sfy its viability
(capability of normal growth and
development) condi$ons while interac$ng in
its environment.
• The team has collec$ve intelligence if the
viability of the group as a whole is required for
each individual to sustain viability.
7. AAractors are leI on the trail
• Foraging is a behavior of
loca$ng food and
transpor$ng it back to the
nest.
• The ants are individuals
responding to their own Deposit a;ractant pheromone on
sensory informa$on and the trail from food.
pheromone signals. Follow the pheromone up its
concentra0on gradient to the
• Pheromones are chemical source.
basis for ant communica$on Increase pheromone concentra0on
to aAract more.
deposited/detected by ants. This posi0ve feedback loop
produces a swarm of ants to quickly
transport the food source.
10. Individual search
Detect the signal
disturbance
Detec$ng aAractor
Selected no$cing
object
Feedback loop
Finding food Following the from the
signal trail environment
Collec$ng and
leaving signal trail Analogy
Increasing pheromone
Modified signal
concentra$on
False pheromone
Fading in $me
11. Quan$ta$ve defini$on of swarm
intelligence
• Performance gains through swarming occurs
when a cri0cal mass of agents come together and
enter a posi0ve feedback loop.
• Explicit use of the environment in agent
interac$ons means that environmental dynamics
are directly integrated into the system’s control,
and in fact can enhance system performance.
12. Narra$ve cues in swarms
• Swarms are communi$es in which decision‐
making takes place based on cues/traces leI
by individual swarm members in the
environment or picked up from their real
ac$vi$es.
• These cues may be small narra$ve or visual
pierces or longer stories.
13. Wri$ng narra$ves as a swarm
Individual search
Detect the signal
disturbance
Dis$nguishing Previous
feature experience
No$cing a story
Detec$ng aAractor Analogy Selected no$cing
object
Feedback loop
Following the from the
signal trail environment
Visibility
Collabora$ng,
Collec$ng and
Wri$ng narra$ve,
cloning the
leaving signal trail
mashing, tagging,
story
geo‐tagging Abduc$on
New
Increasing aAractor Modifying the signal
story
concentra$on
Expanding, transla$ng, interpre$ng
14. Surveillance
• Par$cipa$ng in social networks resides on social
surveillance.
• However, when many transac$ons are
aggregated, pa;erns become visible.
• Narra$ve paAerns may be used to assemble a
detailed profile revealing the ac$ons, habits,
beliefs, loca$ons frequented, social connec$ons,
and preferences of the individual.
• Swarms can use it as environmental informa$on
to adopt them be;er to the environment.
18. Marke$ng swarms by Chuck Brymer
• Today, we are dealing with a swarm where people
gather and deposit informa$on with the collec$ve
intelligence of an en$re social network.
• Ul$mately the swarm decides whether your brand is a
peer or a predator, and does so quickly and
disrup$vely.
• Since you only control part of this informa$on, it will
become more cri$cal than ever to engage the people
who influence swarm communi0es.
• Once a swarm has been launched, human overseers
can observe its emerging behavior and intervene on an
excep0on basis.
19. Untrustworthy communica$on is
possible
• An enemy trying to conceal the search target,
may spread false signals to aAract the agents
to a loca$on of liAle interest.
• Strategy: respond to an external signal only if
it passes a threshold value.
• Strategy: in case of detected communica$on
disturbance enter to an isola0on state for a
$me and act independently not responding to
external signals.
20. Ques$ons
• What kind of aAractors work in the hybrid
ecology swarms?
• Can these aAractors be used for triggering
marke$ng swarms?
• Monitoring swarm paAerns becomes essen$al to
sell beAer, to trigger for swarming, but how can
we monitor automa$cally?
• Can we avoid being monitored without harming
swarming? What and where to restrict access,
and think of security?