ARTIKEL
Communicated by Steven Zucker
Toward Network Intelligence
Alex Pentland
pentland@mit.edu
Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
This article proposes a conceptual framework to guide research in neural
computation by relating it to mathematical progress in other fields and
to examples illustrative of biological networks. The goal is to provide
insight into how biological networks, and possibly large artificial net-
works such as foundation models, transition from analog computation
to an analog approximation of symbolic computation. From the mathe-
matical perspective, I focus on the development of consistent symbolic
representations and optimal policies for action selection within network
settings. From the biological perspective, I give examples of human and
animal social network behavior that may be described using these math-
ematical models.
1 Einführung
Progress in understanding the phenomenon of intelligence as a result of
neural computation has been astounding since the time I was helping
Terry Sejnowski and Steve Zucker, and later Geoffrey Hinton, run work-
shops on computational neuroscience at the Woods Hole Marine Biological
Laboratory. Despite amazing progress, many of the fundamental problems
in understanding network intelligence remain unanswered. Intelligence is
defined by the Oxford English Dictionary as the ability to acquire and apply
knowledge and skills. Heute, most neural computation avoids the hardest
parts of modeling intelligent behavior by framing tasks so that they avoid
problems of representation, Kontext, and intention by having humans se-
lect data that address a specific knowledge or skill task and preestablishing
training or success criteria for the computation.
Extremely large artificial networks such as foundation models, welche
exhibit surprising capacity and flexibility, are rich descriptions of relation-
ships within the data they are trained on because their training is not
supervised. Jedoch, they do not by themselves have the intentionality
or generativity that biological entities require for survival, although they
provide sophisticated perception and example generation capabilities that
could be a major part of an agent that could survive in the real world.
An important question that I consider is whether foundation models have
Neural Computation 35, 525–535 (2023)
https://doi.org/10.1162/neco_a_01536
© 2022 Massachusetts Institute of Technology
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
526
A. Pentland
somehow passed a threshold and possess the abstraction and generativity
capabilities characteristic of symbolic systems.
At an even more general level, there is no “computational theory” in
the sense of Marr and Poggio (1976), who describe the evolutionary fitness
problem that biological networks are solving and connect the specifics of the
network’s computation to solution of the fundamental problems of survival
and reproduction. Mit anderen Worten, we do not understand why biological
neural networks have the structure and behavior that we observe.
In the spirit of exploring possibilities for a computational theory of net-
work intelligence, I focus on three basic issues. The first is how the capa-
bilities of symbolic computation arise, since this is widely considered to be
fundamental to human-level intelligence. Zweite, schnell, stable, and gener-
alizable learning is still something of a mystery; most of today’s methods
require huge numbers of examples that accurately sample the problem dis-
tribution. And third, there is the difficulty that evolution and Darwinian se-
lection require learning actions that improve fitness and survivability rather
than just learning accurate representations or mimicking carefully crafted
training examples. I address each of these three in turn, first by describing
recent mathematical progress in adjacent fields and then presenting exam-
ples of biological networks that illustrate the ideas.
2 From Signals to Symbols
Why do humans have symbolic intelligence? Perhaps the most common hy-
pothesis is that it is to support language and thus social learning. Symbolic
communication requires establishing a shared vocabulary (a set of symbols)
with a clear association to the external world. Unlike the traditional biologi-
cal evolution models where organisms become adapted to their local niches
(and unlike traditional quantization theory, where codebooks are adapted
to local source distributions), language evolution is necessarily a social
phenomenon, since without social interaction, there is no need for shared
vocabularies. Folglich, communication vocabularies must evolve to
balance individual concerns and social exchange. In the neural compu-
tation setting, this means that states of each interacting neural assembly
must have the same environmental associations as the states of connected,
cooperating neurons.
Formally, symbolic representation is based on signal description (z.B.,
quantization), das ist, the assignment of symbols to ranges of sensory sig-
nals. In our 2021 IEEE signal processing paper (Mani, Varshney, & Pentland,
2021), we prove that under surprisingly general conditions, a quantization
network game with cooperative communication will evolve a symbolic rep-
resentation and settle into a Nash equilibrium with local agents having a
shared vocabulary.
Using game theory to characterize this type of longer-range descrip-
tive interaction has long been proposed for understanding how neurons
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Toward Network Intelligence
527
cooperate to segment continuous portions of images (Müller & Zucker,
1991), but the process of generating a symbolic description also requires
that each node chooses a vocabulary for describing the signals that it sees
in such a way that its vocabulary is an accurate description of both the cur-
rent signals and the surrounding nodes’ signals.
In contrast to traditional results in the evolution of symbolic communi-
cation, we found that several vocabularies of symbols may coexist in a Nash
equilibrium of this network learning process. The overlap between vocabu-
laries is high for nodes that communicate frequently and have similar local
sources. This process provides a good account of the emergence of different
languages in separated human groups and the greater frequency of words
specialized to the local situation.
A concrete example is learning the sounds and meanings of words. Unser
2002 Cognition paper (Roy & Pentland, 2002) showed that early word learn-
ing of human infants can be accounted for by quantization of the mutual
information between their audio and visual perceptual streams. Later work
showed that feedback from other people about the quality of audio encod-
ing and appropriateness of referent is what guides the word learning pro-
cess into its more-or-less final form.
Wichtig, these sort of architectures are not specialized just for word
learning. Zum Beispiel, they can be also applied to action selection using
affordance maps, as Gibson (1979), suggested. Zum Beispiel, if red circular
blobs on a green background are associated with a map of affordance such
as “edible,” then outputs of the network can be used to trigger eye move-
ments or further mental processing needed to facilitate picking red berries
on a green bush.
This type of clustering-and-communication feedback may be happening
when we train very large artificial networks by looking at all the relation-
ships within the training corpus. Small networks have long been known
to be quite good at finding local correlations, but in foundation models,
longer-range structure is also very influential, leading to a balance between
local description and long-range comparison, exactly the conditions that
we have shown are sufficient for the emergence of symbolic representa-
tion. The generation of descriptive symbols for heterogeneous phenomena
is a natural and likely common consequence of a cooperative communi-
cation for signal description, and the availability of symbols associated
with environmental regularities pairs nicely with action selection networks
such as the distributed Thompson sampling architecture discussed in
section 4.
3 Rapid, Stable, and Generalizable Learning
Popular neural computation algorithms famously require huge amounts of
Daten, and even specialization of the huge foundation models that are cur-
rently popular typically requires a great deal of data. Auch so, most models
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
528
A. Pentland
exhibit relatively poor generalization, not only because of overfitting but
also because of hidden assumptions, such as the implicit assumption that
data distributions are compact and stable. These are problematic assump-
tions because many social, financial, and biological phenomena have long-
tailed and nonstationary distributions. In recent years the mathematics to
address these limitations have been developed, and we can make use of
these innovations to improve neural computation algorithms.
Zum Beispiel, one of the common frameworks for neural computation is
Q-learning, which maximizes the expected value of the total reward over
successive steps for any finite Markov decision process (FMDP). Q-learning
works to discover an optimal action-selection policy for any given FMDP,
given infinite exploration time and a partly random policy. In der Praxis, Die
convergence of this optimization method is generally very slow. Darüber hinaus,
this sort of reinforcement learning is unstable when a nonlinear function
(such as neural network) is used to represent Q. This instability comes from
the correlations in the sequence of observations and between Q and the tar-
get values, as well as the nonlinearity of Q.
An important advance associated with Q-learning is the ability to con-
nect delayed rewards to actions. The technique of experience replay, a bi-
ologically inspired mechanism that uses a random sample of prior actions
instead of the most recent action, helps suppress spurious correlations, Und
periodic update of Q reduces correlations with the target (Mnih et al., 2015).
This allows Q-learning to be used, zum Beispiel, to develop a play policy for
games like chess or Go.
Q-learning, Jedoch, does not take advantage of recent mathematical re-
sults concerning combination of evidence from different agents (or nodes)
in a network to achieve optimal use of data, speziell, the mathematics
of minimum regret decision making, except for very specific single-agent
exploration policies. I refer to agents rather than nodes in the following to
emphasize the potential for groups of notes to act as action selection alter-
natives. The criterion of minimum regret is the requirement that an agent or
group of agents make the best action selection possible at each time period
given the information and previous experience available at the time.
4 Networks and Thompson Sampling
Classic minimum-regret decision making is called Thompson sampling
(TS), and the literature often refers to this as a bandit problem, because of
the formal equivalence to the question of which slot machine (the “one arm
bandit”) to try in a gambling casino. In the past decade, the mathemati-
cal solutions to such problems have been extended to networks of agents;
Zum Beispiel, a gambler observes the payouts of other casino patrons and
combines those observations with his or her personal knowledge to decide
which slot machine to try next.
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Toward Network Intelligence
529
The classic TS strategy for optimal (minimum regret) decision making
has demonstrated excellent performance in many domains, enough so that
it is a standard approach in domains such as signal processing, medical
decision making, and finance. It shows very fast convergence to optimal
policies, good generalization to new and changing situations (Dubey, Ra-
manathan, Pentland, & Mahajan, 2021), and the ability to work with noisy
and ill-conditioned data inputs.
A fundamental difference between TS and Q-learning is that TS incor-
porates an exploration strategy to help it find the highest reward actions,
whereas Q-learning is an optimization algorithm. This means that TS com-
bines exploitation (optimization by computing the posterior) with explo-
ration (by sampling from the posterior to pick an action or “bandit arm”
that may yield greater rewards). Im Gegensatz, Q-learning simply tells us how
to estimate the Q values but still requires an exploration strategy such as
UCB. In the context of neural networks, typical actions are adjusting net-
work weights.
TS can provide an exploration strategy for Q-learning that has strong op-
timality guarantees. If we were to compute Q-values assuming some para-
metric form of the environment (z.B., gaussian rewards), and then select
actions to explore or exploit based on samples previously drawn from this
parametric form, the resulting strategy could be thought of as being similar
to TS.
This example highlights a second a basic difference between Thompson
sampling and Q-learning. TS typically operates on a discrete, symbolic rep-
resentation of actions, whereas Q-learning is model free. Im Wesentlichen, to com-
pute the posterior reward of an action, TS assumes a parametric model for
the prior distribution and a likelihood function that is updated with obser-
vationen. Q-learning instead starts from zero and simply keeps updating its
parameters without trying to model the environment.
Abschnitte 2 Und 3 and this one have described how cooperative commu-
nication for quantization naturally segments input signals (which can be,
Zum Beispiel, either affordances or percepts) into a set of symbols that can
be used for accurate communication between agents or neurons. The out-
put of such a cooperative quantization computation is exactly the sort of
symbolic parameterization of the environment required for TS to produce
an optimal, minimum-regret action policy.
5 Networks of Smart Neurons
The mathematics of TS can easily be extended to networks of agents, Also
that they may collectively develop an optimal, minimum regret policy for
Aktion. The core of the distributed Thompson sampling (DTS) strategy is for
each decision-making agent to use the experience of other agents to form an
estimate of the prior probability for each potential action and then multiply
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
530
A. Pentland
this prior by each action’s likelihood distribution as determined from their
personal knowledge. This produces a posterior distribution of what reward
each action is expected to produce.
Wichtig, the symbolic representations that naturally emerge from
cooperative signal description can be used as a discrete, symbolic repre-
sentation for DTS. This allows the posterior distribution estimated by DTS
to efficiently and robustly drive the cooperative symbolic encoding. Zum Beispiel-
reichlich, if cooperative signal description distinguishes small red blobs on a
green background, then DTS can efficiently discover that a good reward
can be obtained by using a red blob as a trigger for eye movements or
further mental processing. In this way the encoding for “red blob” can be
mapped from a purely descriptive representation to a functional or inten-
tional representation—for example, an affordance map that relates input
values to possible actions. This is the sort of computation required to im-
plement the automatic releasing-stimuli behaviors seen in simple animals.
Several research groups (including my own) have recently extended DTS
to cover communication-limited networks analogous to the networks seen
in both biological systems and human social behavior (Dubey et al., 2021;
Dubey & Pentland, 2020A, 2020B). We have also extended the DTS frame-
work to adversarial situations and situations where different agents have
different goals. This is accomplished by having each agent compare the
choices he or she makes to the choices of others, and from that comparison
estimate the similarity of the utility function of the other agents to one’s
own utility function, as in Dubey and Pentland (2020A).
These results allow agents to select the best subset of agents with which
to trade experiences and identify agents acting in an adversarial manner or
reacting to different signals. A concrete image example is deciding which
image patches have the same source distribution, thus segmenting the im-
age into regions that are “the same.” Application of these techniques to
neural networks can provide a learning rule for selecting and reinforcing
connections between neurons in order to achieve minimum regret action
selection even in the presence of inhomogeneous input signals and costly
or limited interneuron communication.
In plainer English, these extensions of DTS provide an iterative estima-
tion framework that allows networks of agents or nodes with different in-
puts and different output connections (z.B., different action selection) Zu
modulate their network connectivity in order to maximize their reward
(Dubey & Pentland, 2020A). The result is a minimum regret decision se-
quence for each agent or node and an analogous optimality result for con-
nected subgroups within the group as a whole. Daher, DTS offers a way to
prune large neural networks and discover core sets that are the most impor-
tant for high-reward behaviors. Darüber hinaus, under relatively mild assump-
tionen, the stability of the network and its output can be guaranteed as shown
in our 2020 paper in the Proceedings of the National Academy of Science (Lera,
Pentland, & Sornette, 2020).
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Toward Network Intelligence
531
6 Toward a Computational Theory of Network Intelligence
Developing a computational theory of how intelligence evolved in neural
networks is problematic given our current state of knowledge about the
early evolution of life. Jedoch, we can say much more about the sort of
computational theory that applies to modern biological networks that have
many individual members, such as those of social animals (including ev-
erything from cooperative single-cell organisms to humans), weil das
action selection and reward functions are easier to observe experimentally.
Zum Beispiel, we know that local environments change, and so virtually all
Spezies, from single cells on up, have developed ways of finding new food
sources. The problem of exploring for new food sources while maintaining
sufficient exploitation of current sources is thus a nearly universal problem
for biological networks.
Examples like this exploration-exploitation trade-off can guide the de-
velopment of a computational theory of how biology solves everyday
but critical problems like feeding and reproduction in an uncertain and
changing environment, problems that are certainly an important part of
all species’ survival fitness. Folglich, a natural hypothesis is that the
mathematical patterns we see in today’s biological organisms (einschließlich
primitive single-cell organisms) may be a common networking pattern se-
lected by evolution and thus may plausibly apply to neural networks as
well.
It is encouraging that “distributed Thompson sampling” is a good de-
scription of the group foraging behavior of many social animals (Berger-Tal,
Nathan, Meron, & Saltz, 2014). Group foraging may be viewed as a “portfo-
lio strategy” where the animals do not just choose the maximum posterior
likelihood action (z.B., the action that has been yielding the most food); In-
stead, they make the frequency of different actions proportional to the pos-
terior likelihood. The frequency with which various actions occur can thus
incorporate information from the observable experience of all the animals
in the social group.
Sampling and integration of group experience provide a solution to the
exploration-exploitation dilemma, where animals must allocate some effort
to actions that have been reliably producing good results, while at the same
time exploring for new resources, with the consequence that their portfolio
of frequent actions continuously evolves over time. The fact that this be-
havior is ubiquitous, even in single-cell animals, means that DTS is a good
hypothesis for understanding the computational problems that shaped the
evolution biological neural networks.
7 Human Network Behavior
Studies of human financial decision making, as well as human mobility
and shopping behavior over many days, show the exploration-exploitation
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
532
A. Pentland
pattern characteristic of DTS (Adjodah et al., 2021; Krumme, Lorente,
Cebrian, Moro, & Pentland, 2013). Zum Beispiel, in Adjodah et al. (2021) Wir
found that financial experts use observations about the actions of other ex-
perts to control their portfolio risk in a manner consistent with DTS (Adjo-
dah et al., 2021). An important function of the exploration actions is to avoid
behavioral rigidity, where only a limited number of very familiar actions are
chosen within a group. The phenomenon of insufficient exploration, Ergebnis-
ing in a static portfolio of actions, is familiar in human social networks as
echo chambers and group-think.
Exploration of novel actions is therefore critical for avoiding unfore-
seen risks, finding new opportunities, and adapting to changing conditions.
In the context of neural networks, this means continually experimenting
with weights and pruning unrewarding connections. One of the important
strengths of the DTS framework is that it provides a formal method of deter-
mining when there is insufficient exploration, and thus risk of group-think,
as well as a formula for derisking decisions by accounting for rare outcomes
and sampling error (Dubey & Pentland, 2020B).
Inspired by these mathematical decision systems results, my research
group has worked on the cognitive science version of this decision-making
literature and believe that we have made interesting progress toward relat-
ing human decision making to Thompson sampling style optimal decision
Herstellung. This work began with our 2010 Science paper (Woolley, Chabris,
Pentland, Hashmi, & Malon, 2010), which showed that the collective intel-
ligence of small human groups is strongly correlated with entropy of the
group communication and that it is different from, and often more effec-
tive than, individual human intelligence. This insight was sharpened in our
2013 Scientific Reports paper (Krumme et al., 2013), which showed that the
day-to-day behavior of human populations exhibits the same exploration-
exploitation behavior seen in social animal species.
More recently, unser 2021 Cognition paper (Krafft, Shmueli, Grif-
fiths, Tenenbaum, & Pentland, 2021) showed that commonly observed
individual-level social heuristics closely approximate DTS group decision
making and accurately model human small-group behavior in consumer
financial markets. Unser 2021 Entropy paper (Adjodah et al., 2021) showed
that financial experts show the same DTS behavior and, insbesondere, use
the social information of others to control portfolio risk. These results were
extended by in our 2020 Proceedings of the National Academy of Sciences pa-
pro (Almaatouq et al., 2020), which showed that group performance can be
dramatically improved by selecting agents with similar utility functions,
further motivating the idea of limiting inputs to action selection decisions
via segmentation of the surrounding network by similarity in source distri-
bution and symbolic representation.
We have also seen that the Thompson sampling framework provides
a good model of some important examples of multimodal learning and
decision making. Zum Beispiel, unser 2002 Cognitive Science paper on infant
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Toward Network Intelligence
533
language learning (Roy & Pentland, 2002) showed that audiovisual cues
from adults provided critical cues allowing infants to locate and disam-
biguate nouns within the audio signal. In more recent work, we have shown
that the DTS sampling framework provides a useful model for humans
to create and learn category boundaries in visual stimuli (Epstein, Groh,
Dubey, & Pentland,2021) and for highly efficient domain generalization in
image classification (Dubey et al., 2021).
8 Summary
Neural computation today is dominated by excitement over the new-found
power of deep networks. To move to the next level of performance, the field
will have to address some hard questions about the purpose and nature of
intelligence and how neural computation fits into these larger questions.
This paper makes three main points that address these issues:
1. Cooperative signal description between different node assemblies
naturally generates stable symbolic representations that support ac-
curate, shared communication. This may in part be responsible for
the surprising flexibility of large foundation models.
2. Thompson sampling (TS) can extend Q-learning to symbolic compu-
Station, providing efficient and robust learning of minimum regret ac-
tion policies, including improved network connectivity update rules,
tolerance of erroneous and inhomogeneous inputs, and improved re-
sponse to nonstationary and long-tailed distributions. Daher, TS offers
a principled way to prune large neural networks in order to retain
only the most useful nodes and features.
3. Network minimum regret behavior is very common in real-world bi-
ological networks and systems such as human social networks. It also
appears to apply to social animals more generally, including some
simple and primitive organisms, and thus is plausible as a computa-
tional theory for much of neural computation.
The goal of this article has been to provide a framework to guide research
into how neural computation can produce intelligent behavior. It is my hope
that the research program, scientific papers, experiments and theorems that
I have described here will help future researchers answer these important
Fragen.
Danksagungen
This research is the thesis work of several generations of my graduate stu-
dents, who have been supported by the MIT Media Lab’s Industry Con-
sortium and my Trust Data Consortium within MIT’s Institute for Data
Systems and Society.
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
534
Verweise
A. Pentland
Adjodah, D., Leng, Y., Chong, S., Krafft, P., Moro, E., & Pentland, A. (2021). Accuracy-
risk trade-off due to social learning in crowd-sourced financial predictions. En-
tropy, 23(7), 801. 10.3390/e23070801, PubMed: 34202445
Almaatouq, A., Noriega-Campero, A., Alotaibi, A., Krafft, P., Moussaid, M., & Pent-
Land, A. (2020). Adaptive social networks promote the wisdom of crowds. In
Verfahren der Nationalen Akademie der Wissenschaften, 117(21), 11379–11386. 10.1073/
pnas.1917687117
Berger-Tal, O., Nathan, J., Meron, E., & Saltz, D. (2014). The exploration-exploitation
dilemma: A multidisciplinary framework. PLOS One. https://doi.org/10.1371/
zeitschrift.pone.0095693
Dubey, A., & Pentland, A. (2020A). Private and byzantine-proof cooperative
decision-making. In Proceedings of the 20th Conference on Autonomous Agents and
Multiagent Systems (S. 357–365).
Dubey, A., & Pentland, A. (2020B). Kernel methods for cooperative multi-agent con-
textual bandits. In Proceedings of the International Conference on Machine Learning
(S. 2740–2750).
Dubey, A., Ramanathan, V., Pentland, A., & Mahajan, D. (2021). Adaptive meth-
ods for real-world domain generalization. In Proceedings of the IEEE/CVF Confer-
ence on Computer Vision and Pattern Recognition (S. 14340–14349). Piscataway, NJ:
IEEE.
Epstein, Z., Groh, M., Dubey, A., & Pentland, A. (2021). An experimental study of
social influence and information design. In Proceedings of the ACM on Human-
Computer Interaction (S. 1–18). New York: ACM.
Gibson, J. (1979). The ecological approach to visual perception. Boston: Houghton Mifflin.
Krafft, P., Shmueli, E., Griffiths, T., Tenenbaum, J., & Pentland, A. (2021). Bayesian
collective learning emerges from heuristic social learning. Cognition, 212, 104469.
10.1016/j.cognition.2020.104469, PubMed: 33770743
Krumme, C., Lorente, A., Cebrian, M., Moro, E., & Pentland, A. (2013). The pre-
dictability of consumer visitation patterns. Wissenschaftliche Berichte, 3, 1645. 10.1038/
srep01645, PubMed: 23598917
Lera, S., Pentland, A., & Sornette, D. (2020). Prediction and prevention of dispro-
portionally dominant agents in complex networks. In Proceedings of the National
Akademie der Wissenschaften, 117(44), 27090–27095. 10.1073/pnas.2003632117
Mani, A, Varshney, L., & Pentland, A. (2021). Quantization games on social networks
and language evolution. IEEE Transactions on Signal Processing, 69, 3922–3934.
10.1109/TSP.2021.3090677
Marr, D., & Poggio, T. (1976). From understanding computation to understanding
neural circuitry. AI MIT memo AIM (37). http//hdl.handle.net1721.1/5782
Müller, D., & Zucker, S. (1991) Copositive-plus Lemke algorithm solves polymatrix
games. Operations Research Letters, 10, 285–290. 10.1016/0167-6377(91)90015-H
Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A., Veness, J., Bellemare, M., … Fidje-
Land, Andreas K. (2015). Human-level control through deep reinforcement learn-
ing. Natur, 518(7540), 529–533. 10.1038/nature14236, PubMed: 25719670
Roy, D., & Pentland, A. (2002). Learning words from sights and sounds: A compu-
tational model. Cognitive Science, 26(1), 113–146. 10.1207/s15516709cog2601_4
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Toward Network Intelligence
535
Woolley, A., Chabris, C., Pentland, A., Hashmi, N., & Malone, T. (2010). Evidence for
a collective intelligence factor in the performance of human groups. Wissenschaft, 330
(6004), 686–688. 10.1126/science.1193147, PubMed: 20929725
Received September 10, 2021; accepted January 8, 2022.
l
D
Ö
w
N
Ö
A
D
e
D
F
R
Ö
M
H
T
T
P
:
/
/
D
ich
R
e
C
T
.
M
ich
T
.
/
e
D
u
N
e
C
Ö
A
R
T
ich
C
e
–
P
D
/
l
F
/
/
/
/
3
5
3
5
2
5
2
0
7
1
8
3
0
N
e
C
Ö
_
A
_
0
1
5
3
6
P
D
.
/
F
B
j
G
u
e
S
T
T
Ö
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
PDF Herunterladen