Is it possible to build a machine to do archaeology?
Will this machine be capable of acting like a scientist?
Will this machine be able to understand the way humans acted, or how humans think they acted in the
past?
This book tries to offer some possible answers to these questions and to investigate what it means to
solve “automatically” archaeological problems.
Don’t panic! Even if those questions would have a positive answer, I am not arguing that an artificial
archaeologist will replace human archaeologists, because it will work better and cheaper than we will. We
all know that artificial intelligence will eventually produce robots whose behavior may seem dazzling,
but it will not produce robotic persons. Automatic archaeologists will DO a lot, but they will not BE a
lot. Computational mechanisms cannot carry by themselves the weight of a scientific explanation.
I have tried to create an analogy with an “intelligent” machine, in order to understand the way we think.
We should imagine an automated or artificial archaeologist as a machine able to act as any of us, human
archaeologists, learning through experience to associate archaeological observations to explanations,
and using those associations to solve archaeological problems. It should have its own “cognitive core”
and should interact with the world to make changes or to sense what is happening. In so saying, I am
not arguing that machines run as human brains or that computer representations should be isomorphic
to “mental” states. Rather, I want to understand reasoning processes by understanding the underlying
abstract causal nature behind what archaeologists do. If a computer can be programmed to perform
human-like tasks, it will offer a “model” of the human activity that is less open to argument than the
verbalized explanations that are normal in philosophy. The purpose is then to understand how intelligent
behavior is possible in archaeology.
I am just arguing that the activity of machine and human automata can be described and analyzed in
the same terms. The idea of an intelligent robot should be seen as a model of archaeologist’s behavior
rather than an explanation of his or her mind. Computer hardware and programming techniques enable
the model builder to construct virtual creatures that behave in intelligent and flexible ways under natural
conditions. They provide powerful (and perhaps indispensable) tools for building such creatures, but
they can play no role as explanatory kinds by themselves.
In some way, computational intelligence provides social scientists with a set of tools with the same
degree of finesse as those used in current qualitative studies and with the same mobility, the same
capacities of aggregation and synthesis, as those used in quantitative studies by other social sciences.
The limitations of these tools and methods are the same as those of any instrument from any scientific
discipline. Instead of being restricted to the usual representational schemes based on formal logic and
ordinary language, computational approaches to the structure of archaeological reasoning can include
many useful representations such as prototypical concepts, concept hierarchies, conceptual spaces, production
rules, associative memories, causal networks, mental images, and so on. Researchers concerned
with the growth of scientific knowledge from a computational perspective can go beyond the narrow
resources of inductive logic to consider algorithms for generating quantitative laws, discovering causal
relationships, forming concepts and hypotheses, and evaluating competing explanatory theories. This
book presents tools and methods that liberate us from the narrow constraints of words by enforcing rigor
in a non-classical way, namely via the constraint of computational realizability
Maybe some of you will say that we do “not yet” have automatic archaeologists, but we should hurry
up to the engineering department and build them for having someone able to substitute us in the tedious
task of studying ourselves and our past. Other readers will claim: “fortunately, such a machine will never
exist!” “Why we need such an awful junk? Computers cannot emulate humans.” These critics seem to
think that computer programs are guilty of excessive simplification, of forcing knowledge, or distorting
it, and of failing to exploit fully the knowledge of the expert, but it seems to me that it is archaeology,
and not computer programs, what is “narrow minded.” The saddest thing is that archaeologists do not
know how they know archaeological matters.
The so called “intelligent” machines incite instinctive fear and anger by resembling ancestral threats, a
rival for our social position as more or less respected specialists. But robots are here, around us. I have
never heard of a claim against washing machines selecting “intelligently” the best way to wash a specific
tissue, or a photo camera with an “intelligent” device measuring luminance and deciding by itself the
parameters to take the picture. So, why have fear of a machine classifying a prehistoric tool and deciding
“intelligently” its origin, function and/or chronology? Rather than arguing whether a particular behavior
should be called intelligent or not,a point that is always debatable, I try to provide answers to the following
question: Given some behavior that we find interesting in some ways, how does the behavior come
about? Rather than use intuition as the sole guide for formulating explanations of past human behavior,
we need a theory of why a specific computation or a group of related computations should be performed
by a system that has certain abilities.
The discussion is between what is considered an artificial way of reasoning (computer programs), and
a natural way of reasoning (verbal narrative). Critics of computationalism insist that we should not
confound scientific statements with predicate logic operations, since discursive practices or argumentations
observed in a scientific text are not “formal.” By that reason, they are tributary, to a certain extent,
from the Natural Language and the narrative structure (literary) of which scientific texts derive. I take
the opposite approach: scientific problem solving stems from the acquisition of knowledge from a
specific environment, the manipulation of such knowledge, and the intervention in the real world with
the manipulated knowledge. The more exhaustive and better structured the knowledge base, the more
it emulates a scientific theory and the easier will be the solution to the scientific problem, and more
adequate the interpretations we get.
My personal approach is based on a fact that archaeologist could not evaluate 15 years ago: computer
programs do work in real science, not only in archaeology. Maybe they are more successful in other
“harder” sciences, but we cannot deduce from this fact that archaeology is a different kind of science. We
should instead rebuild archaeology. Simulating or reproducing the way archaeologists think today is not
the guide to understand archaeology, because we are doing archaeology in the wrong way! Computable
archaeology, if you do not like the expression “automatic archaeology,” is the proper way of exploring
new ways of thinking old concepts.
In other scientific domains the performance of humans at a particular task has been used to design a
robot that can do the same task in the same manner (and as well). In many different domains it has been
shown how ‘robot scientists’ can interpret experiments without any human help. Such robots generate a
set of hypotheses from what it is known about a scientific domain, and then design experiments to test
them. That is, a robot scientist can formulate theories, carry out experiments and interpret results. For
instance, the robot biochemist developed by Ross King of the University of Wales at Aberystwyth, and
his colleagues, does everything a flesh-and-blood scientist does—or, rather, it does what philosophers
of science say that scientists ought to do. That is, it formulates hypotheses from observations, conducts
experiments to test them, and then formulates new hypotheses from the results. And, it does so as effectively
as a person. The intellectual input comes from deciding, on the basis of the results obtained,
which experiments to do next until you have filled in all the blanks. The robot scientist was able to do
this. It was fitted with artificial intelligence software that could perform the logical processes involved
in making such decisions, and this software was given a representation of the pathway chosen (one of
those by which amino acids, the building blocks of proteins, are made) from which to work. The robot
scientist can infer hypotheses to explain observations, infer experiments that will discriminate between
these hypotheses, actually do the experiments and understand the results.
Consequently, the design of an automated archaeologist should not be considered a mere science fiction
tale. It is a technological reality. Research in cognitive robotics is concerned with endowing robots
and software agents with higher level cognitive functions that enable them to reason, act and perceive
in changing, incompletely known, and unpredictable environments. Such robots must, for example, be
able to reason about goals, actions, when to perceive and what to look for, the cognitive states of other
agents, time, collaborative task execution, and so forth. In short, cognitive robotics is concerned with
integrating reasoning, perception and action within a uniform theoretical and implementation framework.
The question of whether it is possible to such machines to automate the scientific process should be of
both great theoretical interest and increasing practical importance because, in many scientific areas, data
are being generated much faster than they can be effectively analyzed.
The book is divided into four parts. The first one introduces the subject of “artificial intelligence” within
the apparently restricted domain of archaeology and historical sciences. This introductory part contains
two chapters. “‘Automatic’” Archaeology: a useless endeavor, an impossible dream, or reality?” The
first provides an overview of the approach. After discussing the basic concepts of automata theory, the
first elements of a formalization of archaeological reasoning are presented. The very idea of archaeological
problems is introduced from the point of view of cause-effect analysis and social activity theory.
The relationship between archaeological, anthropological, and historical problems is studied in detail,
to serve as a basis for a presentation of how a mechanical problem solving procedure would look like
in those domains. The chapter ends with a very short presentation of the diversity in current Artificial
Intelligence theory and techniques.
The second chapter, “Problem Solving in the Brain and by the Machine,” presents the classical artificial
intelligence approach to problem solving as search and planning. Rule-based systems are discussed,
focusing in its philosophical foundations. Jean Claude Gardin’s logicist analysis is used as a relevant
archaeological example, together with some of the current expert systems used in practical archaeology.
A final debate leads the reader to a discussion about “rationality” and the shortcomings of traditional
artificial intelligence and expert systems.
The second part of the book is the most technical one and presents a detailed but understandable account
of learning algorithms and neural networks. It has been divided into two chapters. The third chapter,
“Computer Systems that learn,” develops the criticism of the classical approach to “intelligent robotics,”
presenting the way computer systems and “intelligent” robots may learn. Learning is here presented as
a predictive task that can be simulated by computers. Many archaeological cases are used through this
chapter to understand the algorithmic nature of experimentation and discovery tasks.
The fourth chapter, An introduction to Neurocomputing offers a presentation of neural networks. After
discussing in plain language what neural networks are, some algorithms are introduced with a minimum
of mathematical jargon, here reduced to the basic arithmetic operations. Backpropagation networks are
exhaustively analyzed, together with radial basis functions, self-organized maps, Hopfield networks,
and other advanced architectures.
Part III constitutes the core of the book, and discusses different examples of computational intelligence
in archaeology, with cases concerning rock-art, lithic tools, archeozoology, pottery analysis, remote
sensing, ancient settlement investigation, funerary ritual, social organization in prehistoric societies,
etc. It has been divided into six chapters.
In chapter 5, Visual and Non-Visual Analysis in Archaeology some of the elements introduced in chapter
1 are developed. A general approach towards an “intelligent” pattern recognition system is presented,
discussing the differences between a true visually based system and another one, which uses identified
previously—instead of visual—data. This chapter serves as an introduction to the following ones, where
practical and relevant examples of archaeological neurocomputing are shown in the domains of shape,
texture, composition, spatiotemporal and functional analysis.
Chapter 6, Shape Analysis in Archaeology defines the concept of “shape” and presents different approaches
to shape representation, analysis, and interpretation. Emphasis has been placed on the analysis
of three-dimensional objects and the study of complex shapes.
Chapter 7, Texture and Compositional Analysis in Archaeology defines the concepts of “texture”
and “composition.” It also presents many archaeological applications of neurocomputing in these domains.
Chapter 8, Spatiotemporal Analysis in Archaeology has been written in order to explain the way spatial
and temporal data (frequencies and densities of archaeological findings, for instance) can be analyzed
using neural networks and other similar technologies. The spatial interpolation problem is posed, and
different methods for finding a solution are evaluated, showing many real examples. Remote sensing also
finds its place in this chapter. Time series and chronological problems are also a form of interpolation
problem. Neural networks can be used to solve it, but we also need specifically organized networks to
deal with recursiveness and related questions. The focus is on spatiotemporal explanatory models, not
only from a strictly archaeological point of view but with a more general social science and historical
perspective.
In Chapter 9, An Automated Approach to Historical and Social Explanation visually based explanatory
approaches are substituted by a more general account of simulation and modeling, which illustrates how
social processes can be simulated as computational mechanisms to be understood. The idea of social
classification is discussed, and many examples of simulating social interaction using “populations” of
computer programs are finally presented.
To conclude our journey into the automatization of scientific reasoning, the book ends with a Part IV that
presents a theoretical discussion on the philosophy of social sciences and the benefits of computers and
nonlinear algorithmic approaches. This part is composed of a single chapter that explores the theoretical
consequences that may arise when using computational intelligence technologies to study the human
past. Here the “robot” analogy gives its place to a proper account of a Computational Philosophy of
Archaeology and related sciences.
It is important to take into account that this is a book on “computational intelligence” in archaeology,
and not on “computer applications in archaeology.” I have focused the text on the very concept of “explanation,”
and what it really means to explain archaeological (and historical) data. Therefore, important
and fashionable concepts that are not properly related to “explanation” have less relevance. The reader
may ask why I have not included more references to fashionable and apparently modern issues like geoxvi
graphic information systems, visualization and virtual reality. The answer is that these subjects appear
in the book, but in a different envelope, insisting in their contributions to archaeological explanation.
Therefore, GIS techniques have been included in chapter 8 on spatiotemporal explanation, and all the
discussion on virtual reconstructions has a more logical place in chapter 6 on shape analysis, but it is
also analyzed in chapter 10. The reader is referred to other books for the practical side of data bases,
GIS, CAD and visualization software. This is a book on the interface between technique and theory.
Although some “how-to” is presented, and many practical applications are referred, the book merely
opens a door, encouraging the reader to begin a research along this line.
Do not look for a classic presentation of the archaeological practice. This is an unconventional book with
very little respect for tradition. In a first reading, the text may seem highly skewed towards computational
intelligence, with very little traditional archaeological stuff. Even the number of traditional archaeological
references is surprisingly small. This is because my goal has been to open new grounds in archaeology
and the social sciences. Technology is not the solution, but it is the way we have to follow if we want
to rethink the way archaeology has been done. This emphasis on new ways to understand ancient times
explains the apparently minor relevance of traditional aspects. However, they are not absent. They have
acquired a new appearance, as a careful reading will prove.
This is not an encyclopedia of archaeological methods and explanations. I could not present all
aspects of the archaeological research process nor all available computer science methods. Because
any book needs to be focused, I have had to obviate many important aspects that in other circumstances
would be interesting. If a majority of readers find the book relevant, and I have the chance to do more
research work in this “computable” archaeology, new chapters on archaeological site formation processes
or intelligent virtual archaeology environments will follow. The technology is evolving, and each day
sees some new advancement. For all information that couldn’t be included in the book, and for periodic
updates of theories, techniques and technologies, the reader is referred to its related Web page:
http://antalya.uab.es/prehistoria/Barcelo/IGIBook.html