Authorea

Joe Corneli add some outtakes about 9 years ago

Commit id: f781f74a0e280f01848c3f3bf666f42536de1ea4

deletions | additions

% What is the goal of the computation (input and output) % Why is it appropriate (formal spec e.g. considering externalities) % what is the logic of the strategy by which it can be carried out. \textbf{[Do we need to include the partial repetition below or is the above formal enough? Could these bulleted ideas be condensed into one or two paragraphs]} Following Section \ref{specs-overview}, the 13 criteria can be used to evaluate the serendipity potential of existing systems, which we discuss first. The 13 criteria could also be used to guide design of future systems to maximise potential for serendipity; we explore this in the thought experiment outlined in Section \ref{sec:ww}. \subsubsection*{Key condition for serendipity} \begin{itemize} \item \textbf{Focus shift}: A focus shift is linked to re-evaluation of data, processes, or products. It may precipitate changes in the entire framework of evaluation or its effects may be more contained. Such reevaluation could be modelled using a multi-agent architecture, in which each agent has a goal and evaluates generated products relative this goal, but in which agents also share their products with other, who then evaluate them against their own metrics. \end{itemize} \subsubsection*{Components of serendipity} \begin{itemize} \item \textbf{Prepared mind}: This comprises the background knowledge, unsolved problems, current goal, programming, and operating environment of a computational system. %% \item \textbf{Serendipity trigger}: The generation or observation of a potentially novel example, concept, or conjecture, etc., which precedes a discovery in a computational system.\footnote{Triggers are often examples without an explanation, rather than wholly-formed concepts.} The trigger is outside of the direct control of the system components responsible for evaluations. %% \item \textbf{Bridge}: Reasoning and/or programmatic interaction brings about a focus shift at an opportune juncture, building on prior preparation and on the serendipity trigger. The bridge may be constructed on the basis of logical methods, analogies, conceptual blending, evolutionary search, automated theory formation and may draw on interactions with other systems. %% \item \textbf{Result}: The discovery itself may be a new product, artefact, process, hypothesis, use for an object, etc., generated by computational means, which may influence the future operations of the system. \end{itemize} \subsubsection*{Dimensions of serendipity} \begin{itemize} \item \textbf{Chance}: Controlled randomness in AI systems is well-established, e.g. in Genetic Algorithms and search. Chance also applies in connection with an under-determined outside world (see below). %% \item \textbf{Curiosity}: The system needs to expend discretionary computational effort on the serendipity trigger. This may be accompanied by system features that an observer would describe as playfulness, inventiveness, and the drive to experiment or understand. %% \item \textbf{Sagacity}: Sagacity be modelled by employing reasoning over multiple application domains simultaneously; or, again, with a social analogue in cases where the system does not know, but ``knows who to ask.'' %% \item \textbf{Value}: The result should be interesting or useful, as judged by the system, the programmer, the user, or another party (potentially another system). \end{itemize} \subsubsection*{Environmental factors} \begin{itemize} \item \textbf{Dynamic world}: Connections with other systems, data sources, or user input, e.g., via the web, which is highly dynamic -- or in the context of a larger simulation. %% \item \textbf{Multiple contexts}: Reasoning which operates across domains, such as analogical reasoning, or that considers multiple perspectives, as in systems with social awareness. %% \item \textbf{Multiple tasks}: Multiple goals or targets that compete for resources. The system may be implemented using a multithreaded, parallel processing design. %% \item \textbf{Multiple influences}: This may again be modelled as a multi-agent systems, as or multiple interacting systems, each with different knowledge and goals. The source of unexpectedness may be arise on various levels, and a system may bring this to bear using techniques of reflection. \end{itemize}

Van Andel's \citeyear{van1994anatomy} ``patterns of serendipity'' are often instances of this broader pattern. %% \begin{quote} %% ``Given research set up for a certain purpose, some unexpected, puzzling data, and a scientist capable of being puzzled -- given all of these, an accidental discovery will occur, because the relationship between fact and theory in science is such that it must occur.'' %% \end{quote} In future work, we would like to explore the usefulness of the somewhat more formal theory of \emph{design patterns} \cite{alexander1999origins} for designing with serendipity in mind. Alexandrian design patterns are by no means limited to computing: the approach has its origins in architecture and urban planning \cite{alexander1979timeless,alexander1977pattern}. Design patterns prescribe and describe: they provide models \emph{for} as well as models \emph{of} \cite[p. 93]{geertz1973interpretation}. Thus, when Alexander describes the pattern \emph{A place to wait}, he is also telling readers that it may be a good idea to consider building a place to wait when designing a living environment. In connection with our understanding of serendipity as closely associated with deviations from familiar patterns (in the everyday sense of the word), it is interesting to ask how it can play a role in the creation of new design patterns and pattern languages. Noticing and describing a new pattern is almost the antithesis of ``pattern recognition'' in the usual computing sense. As a beginning, we examined the 14 ``patterns of serendipity'' selected and described by van Andel \citeyear{van1994anatomy}, using the criteria described in our Section \ref{sec:connections-to-formal-definition}. We found all of these patterns do indeed include a focus shift, a prepared mind, a serendipity trigger, a bridge, and a result, although two of the patterns raised questions: \begin{itemize} \item In the case of \emph{Testing popular belief}, van Andel focuses on an account of a medical practise that originated in a folk claim, namely cowpox-derived immunity to smallpox. It is challenging in this case to identify one specific serendipity trigger -- although a curious chain of events connected Edward Jenner with the smallpox vaccine. It may be most appropriate to think of Jenner himself as the serendipity trigger at the societal level: his ``relentless promotion and devoted research of vaccination \ldots changed the way medicine was practised'' \cite{riedel2005edward}. %% This effect, for milkmaids, might %% indeed be called serendipitous. Indeed, the medical use of cowpox has %% been described as ``widely know'' \cite{riedel2005edward} prior to its %% popularisation by Edward Jenner. Nevertheless, Jenner's \item \emph{Inversion} is closer to what is called an \emph{antipattern} in the design pattern literature \cite{brown1998antipatterns}. Van Andel describes the story of a researcher observing an effect due to the anticoagulant heparine which was precisely the opposite of the one sought -- factors that \emph{cause} blood clotting -- and then failing to acknowledge that this observation was important for over 40 years. The result was eventually seen to be of value, however, again, this may be a pattern of \emph{antiserendipity}. \end{itemize} Among the 14 patterns, four are cases of ``perfect'' serendipity from the point of view of our extended set of criteria (i.e.~they included all of the 13 components, dimensions, and environmental factors) -- these patterns were \emph{Successful error}, \emph{Side effect}, \emph{Wrong hypothesis}, and \emph{Outsider}. % We wondered whether these were patterns might be used to support serendipity in other settings -- such as the Writers Workshop. Table \ref{tab:reinterpret} gives an initial sketch, and initial experiments that will bring this material to computational life are underway. \begin{table}[p] \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Successful error}} & \\ \emph{Van Andel's example}: & Post-it\texttrademark\ notes \\[.2cm] {\tt presentation}& Systems should be prepared to share interesting ideas even if they don't know directly how they will be useful. \\ {\tt listening} & Systems should listen with interest, too. \\ {\tt feedback} & Even interesting ideas may not be ``marketable.''\\ {\tt questions} & How is your suggestion useful? \\ {\tt reflections} & New combinations of ideas take a long time to realise, and many different ideas may need to be combined in order to come up with something useful.\\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Side effect}} & \\ \emph{Van Andel's example}: & Nicotinamide used to treat side-effects of radiation therapy proves efficacious against tuberculosis. \\[.2cm] {\tt presentation}& Systems should use their presentation as an experiment. \\ {\tt listening} & Listeners should allow themselves to be affected by what they are hearing. \\ {\tt feedback} & Feedback should convey the nature of the effect.\\ {\tt questions} & The presenter may need to ask follow-up questions to gain insight. \\ {\tt reflections} & Form a new hypothesis before seeking a new audience. \\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Wrong hypothesis}} & \\ \emph{Van Andel's example}: & Lithium, used in a control study, had an unexpected calming effect. \\[.2cm] {\tt presentation}& How is this presentation interpretable as a (``natural'') control study? \\ {\tt listening} & Listeners are ``guinea pigs''.\\ {\tt feedback} & Discuss side-effects that do not necessarily correspond to the author's perceived intent. \\ {\tt questions} & Zero in on the most interesting part of the conversation.\\ {\tt reflections} & Revise hypotheses to correspond to the most surprising feedback. \\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Outsider}} & \\ \emph{Van Andel's example}: & A mother suggests a new hypothesis to a doctor. \\[.2cm] {\tt presentation}& The presenter is here to learn from the audience. \\ {\tt listening} & The audience is here to give help, but also to get help.\\ {\tt feedback} & Feedback will inevitably draw on previous experiences and ideas.\\ {\tt questions} & What is the basis for that remark?\\ {\tt reflections} & How can I implement the suggestions?\\ \end{tabular} \caption{Reinterpreting patterns of serendipity for use in a computational poetry workshop\label{tab:reinterpret}} \end{table}

The real problem with computers is not that they only do what they're told, but that the act of programming forces us to confront the emergence of the new \cite{mead1932philosophy}. % Indeterminacy forms an important part of any proposal for ``intelligent machines'', after Turing: \begin{quote} ``\emph{They will make mistakes at times, and at times they may make new and very interesting statements, and on the whole the output of them will be worth attention to the same sort of extent as the output of a human mind}.'' \cite{turing-intelligent} \end{quote} Serendipity has played a role in the large-scale history of the computing field \cite{de2013turing} and in artistic applications of computer technology \cite{reichardt1969cybernetic}. We aim to clarify the role it has to play in the future development of computational creativity.

It is worth noting that current systems in computational creativity, almost as a rule, do \emph{not} consume or evaluate the work of other systems.\footnote{An exception that proves the rule is Mike Cook's {\sf AppreciationBot}, which is a reactive automaton that is solely designed to ``appreciate'' tweets from {\sf MuseumBot}; see \url{https://twitter.com/AppreciationBot}.} Developing systems that could successfully navigate this collaborative exercise would be a significant advance in the field of computational creativity. Since the experience is about \emph{learning} rather than winning, there is little motivation to ``game the system'' \cite{lenat1983eurisko}.

\section{Literature review} \label{sec:literature-review} In this section, we give a short overview covering the etymology of the term ``serendipity'' and trace its development in order to pin down the key commonalities from many definitions and instances. In particular, we point out key conditions of serendipity, their components and general characteristics, including environmental factors. The structure of this section follows and updates an earlier survey from \citeA{pease2013discussion}, drawing connections with the new formal model described above. \subsection{Etymology and selected definitions} \label{sec:overview-serendipity} The English term ``serendipity'' derives from the 1302 long poem \emph{Eight Paradises}, written in Persian by the Sufi poet Am\={\i}r Khusrow in Uttar Pradesh.\footnote{\url{http://en.wikipedia.org/wiki/Hasht-Bihisht}} In the English-speaking world, its first chapter became known as ``The Three Princes of Serendip'', where ``Serendip'' represents the Old Tamil-Malayalam word for Sri Lanka (%{\tam சேரன்தீவு}, \emph{Cerantivu}), ``island of the Ceran kings.'' The term ``serendipity'' is first found in a 1757 letter by Horace Walpole to Horace Mann: \begin{quote} \emph{``This discovery is almost of that kind which I call serendipity, a very expressive word} \ldots \emph{You will understand it better by the derivation than by the definition. I once read a silly fairy tale, called The Three Princes of Serendip: as their Highness travelled, they were always making discoveries, by accidents \& sagacity, of things which they were not in quest of}[.]''~\cite[p. 633]{van1994anatomy} \end{quote} The term became more widely known in the 1940s through studies of serendipity as a factor in scientific discovery, surveyed by Robert Merton and Elinor Barber \citeyear{merton} in their 1957 analyis ``The Travels and Adventures of Serendipity, A Study in Historical Semantics and the Sociology of Sciences''. Merton and Barber define the term as follows: \begin{quote} \emph{``The serendipity pattern refers to the fairly common experience of observing an unanticipated, anomalous and strategic datum which becomes the occasion for developing a new theory or for extending an existing theory.''} \cite[p. 635]{van1994anatomy} \end{quote} In 1986, Philippe Qu\'eau described serendipity as ``the art of finding what we are not looking for by looking for what we are not finding'' \cite{eloge-de-la-simulation}, as quoted in \cite[p. 121]{Campos2002}. Pek van Andel \citeyear[p. 631]{van1994anatomy} describes it simply as ``the art of making an unsought finding''. Roberts \citeyear[pp. 246--249]{roberts} records 30 entries for the term ``serendipity'' from English language dictionaries dating from 1909 to 1989. % Classic definitions require the investigator not to be aware of the problem they serendipitously solve, but this criterion has largely dropped from dictionary definitions. Only 5 of Roberts' collected definitions explicitly say ``not sought for.'' Roberts characterises ``sought findings'' in which an accident leads to a discovery with the term \emph{pseudoserendipity} \cite{chumaceiro1995serendipity}. % While Walpole initially described serendipity as an event, it has since been reconceptualised as a psychological attribute, a matter of sagacity on the part of the discoverer: a ``gift'' or ``faculty'' more than a ``state of mind.'' Only one of the collected definitions, from 1952, defined it solely as an event, while five define it as both event and attribute. However, there are numerous examples that exhibit features of serendipity which develop on a social scale rather than an individual scale. For instance, between Spencer Silver's creation of high-tack, low-adhesion glue in 1968, the invention of a sticky bookmark in 1973, and the eventual launch of the distinctive canary yellow re-stickable notes in 1980, there were many opportunities for Post-its\texttrademark\ \emph{not} to have come to be \cite{tce-postits}. Accordingly, Merton and Barber argue that the psychological perspective needs to be integrated with a \emph{sociological} one.\footnote{ ``For if chance favours prepared minds, it particularly favours those at work in microenvironments that make for unanticipated sociocognitive interactions between those prepared minds. These may be described as serendipitous sociocognitive microenvironments'' \cite[p. 259--260]{merton}.} Large-scale scientific and technical projects generally rely on the ``convergence of interests of several key actors'' \cite{companions-in-geography}, along with other supporting cultural factors. Umberto Eco \citeyear{eco2013serendipities} focuses on the historical role of serendipitous mistakes and falsehoods in the production of knowledge. It is important to note that serendipity is usually discussed within the context of \emph{discovery}, rather than \emph{creativity}, although in typical parlance these terms are closely related \cite{jordanous12jims}. In our definition of serendipity, we have made use of Henri Bergson's distinction: \begin{quote} ``\emph{Discovery, or uncovering, has to do with what already exists, actually or virtually; it was therefore certain to happen sooner or later. Invention gives being to what did not exist; it might never have happened.}''~\cite{bergson2010creative} \end{quote} As we have indicated serendipity would seem to require features of both; that is, the discovery of something unexpected and the invention of an application for the same. We must complement \emph{analysis} with \emph{synthesis} \cite{delanda1993virtual}. The balance between these two features will differ from case to case. In the next subsection we will review several historical examples. First, one further point should be made with reference to the ``The Three Princes of Serendip''. Prior to Walpole's coinage, this story had been adapted by Voltaire into an early chapter of \emph{Zadig}, and in turn ``the method of Zadig'' informed subsequent approaches both to fiction writing and natural science. This method is rooted firstly in discovery: \begin{quote} ``[Zadig] \emph{pry’d into the Nature and Properties of Animals and Plants, and soon, by his strict and repeated Enquiries, he was capable of discerning a Thousand Variations in visible Objects, that others, less curious, imagin’d were all alike.}''~\cite[pp. 21--22]{zadig} \end{quote} \noindent Secondly, from disparate observations, Zadig is often able to assemble a coherent picture: \begin{quote} \emph{It was his peculiar Talent to render Truth as obvious as possible: Whereas most Men study to render it intricate and obscure.}~\cite[p. 54]{zadig} \end{quote} Similarly, but in reverse, a coherent picture may be reduced to fragmented pieces each of which may tell a very different story from the whole. This is illustrated in Zadig's misadventure with a broken tablet, in which one fragment of a poem of praise reads as treasonous provocation. In describing the various features of serendipity below, we will draw connections with the schematic diagram presented in Section \ref{specs-overview}, in order to unfold the multifaceted notion of serendipity. \subsection{Connecting our formal definition to literature} \label{sec:connections-to-formal-definition} \subsubsection*{Key condition for serendipity} \begin{itemize} \item \textbf{Focus shift}: ``\emph{After removing several of the burdock burrs (seeds) that kept sticking to his clothes and his dog's fur,}~[de Mestral]~\emph{became curious as to how it worked. He examined them under a microscope, and noted hundreds of `hooks' that caught on anything with a loop, such as clothing, animal fur, or hair. He saw the possibility of binding two materials reversibly in a simple fashion, if he could figure out how to duplicate the hooks and loops.}''~\cite{wiki:velcro} % \inlineitem{This corresponds to the identification of $T^\star$, which is common to both sides of the diagram. \citeA{creativity-crisis} write that: ``To be creative requires divergent thinking (generating many unique ideas) and then convergent thinking (combining those ideas into the best result).'' Accordingly $T^\star$ may be thought of as an evolving vector of interesting possibilities or ``strategic data'' \cite[p. 507]{merton1948bearing}. In de Mestral's case, the initial idea of a hook-and-loop fastener occurred in 1941 -- followed by a full decade of experimentation before he was ready to file a patent claim. } \end{itemize} \subsubsection*{Components of serendipity} \begin{itemize} \item \textbf{Prepared mind}: Fleming's ``prepared mind'' included his focus on carrying out experiments to investigate influenza as well as his previous experience that foreign substances in petri dishes can kill bacteria. He was concerned above all with the question ``Is there a substance which is harmful to harmful bacteria but harmless to human tissue?'' \cite[p. 161]{roberts}. %% % \inlineitem{This corresponds to the prior training $p$ and $p^{\prime}$ in our diagram.} \item \textbf{Serendipity trigger}: The trigger does not directly cause the outcome, but rather, inspires a new insight. It was long known by Quechua medics that cinchona bark stops shivering. In particular, it worked well to stop shivering in malaria patients, as was observed when malarial Europeans first arrived in Peru. The joint appearance of shivering Europeans and a South American remedy was the trigger. That an extract from cinchona bark can cure and can even prevent malaria was subsequently revealed. % \inlineitem{This corresponds to the stimulus $T$ in our diagram.} %% \item \textbf{Bridge}: These include reasoning techniques, such as abductive inference (what might cause a clear patch in a petri dish?); analogical reasoning (de Mestral constructed a target domain from the source domain of burs hooked onto fabric); and conceptual blending (Kekul\'e blended his knowledge of molecule structure with his vision of a snake biting its tail). The bridge may also rely on new social arrangements, such as the formation of cross-cultural research networks. % \inlineitem{This corresponds to the actions based on $p^{\prime}$ taken on $T^\star$ leading to $R$.} %% \item \textbf{Result}: This may be a new product, artefact, process, hypothesis, a new use for a material substance, and so on. The outcome may contribute evidence in support of a known hypothesis, or a solution to a known problem. Alternatively, the result may itself be a {\em new} hypothesis or problem. The result may be a ``pseudoserendipitous'' in the sense that it was {\em sought}, while nevertheless arising from an unknown, unlikely, coincidental or unexpected source. More classically, it is an \emph{unsought} finding, such as the discovery of the Rosetta stone. % \inlineitem{This corresponds to our $R$. Note that $R$ may imply updates to $p$ or $p^{\prime}$ in further phases of research.} \end{itemize} \subsubsection*{Dimensions of serendipity} Whereas the foregoing items are the central features of the definition, the following further characterise the circumstances under which serendipity occurs in practice. \begin{itemize} \item \textbf{Chance}: Fleming \citeyear{fleming} noted: ``There are thousands of different moulds'' -- and ``that chance put the mould in the right spot at the right time was like winning the Irish sweep.'' % \inlineitem{One must assume that relatively few triggers $T^\star$ that are identified as interesting actually lead to useful results; in other words, the process is fallible.} %% \item \textbf{Curiosity}: Venkatesh Rao \citeyear{rao2011tempo} refers to a \emph{cheap trick} that takes place early on in a narrative in order to establish the preliminary conditions of order. Curiosity with can play this role, and can dispose a creative person to begin, or to continue, a search into unfamiliar territory. % \inlineitem{The prior training $p$ causes interesting features to be extracted, even if they are not necessarily useful; $p^{\prime}$ asks how these features \emph{might} be useful. } %% \item \textbf{Sagacity}: This old-fashioned word is related to ``wisdom,'' ``insight,'' and especially to ``taste'' -- and describes the attributes, or skill, of the discoverer that contribute to forming the bridge between the trigger and the result. In many cases, such as an entanglement with cockle-burs, many others will have already been in a similar position and not obtained an interesting result. Once a phenomenon has been identified as interesting, the disposition of the investigator may lead to a dogged pursuit of a useful application or improvement. % \inlineitem{Rather than a simple look-up rule, $p^{\prime}$ involves creating new knowledge.} %% \item \textbf{Value}: Note that the chance ``discovery'' of, say, a \pounds 10 note may be seen as happy by the person who finds it, whereas the loss of the same note would generally be regarded as unhappy. Positive judgements of serendipity by a third party would be less likely in scenarios in which ``One man's loss is another man's gain'' than in scenarios where ``One man's trash is another man's treasure.'' If possible we prefer this sort of independent judgement \cite{jordanous:12}. % \inlineitem{The evaluation $|R|>0$ may be carried out ``locally'' (as an embedded part of the process of invention of $R$) or ``globally'' (i.e.~as an external process). } \end{itemize} \subsubsection*{Environmental factors} \begin{itemize} \item \textbf{Dynamic world}: Information about the world develops over time, and is not presented as a complete, consistent whole. In particular, value may come later. Van Andel \citeyear[p. 643]{van1994anatomy} estimates that in twenty percent of innovations ``something was discovered before there was a demand for it.'' % \inlineitem{$T$ (and $T^\star$) appears within a stream of data with indeterminacy. There is a further feedback loop, insofar as products $R$ influence the future state.} %% \item \textbf{Multiple contexts}: One of the dynamical aspects at play may be the discoverer going back and forth between different contexts, with different stimuli. 3M employee Arthur Fry sang in a church choir and needed a good way to mark pages in his hymn book; he happened to have been attending seminars offered by his colleague Silver about restickable glue. % \inlineitem{This is reflected directly in our model by the difference between the ``context of discovery'' involving prior preparations $p$, and the ``context of invention'' involving prior preparations $p^{\prime}$. Both of these may be subdivided further.} %% \item \textbf{Multiple tasks}: Even within what would typically be seen as a single context, a discoverer may take on multiple tasks that segment the context into sub-contexts, or that cause the investigator to look in more than one direction. The tasks may have an interesting \emph{overlap}, or they may point to a \emph{gap} in knowledge. As an example of the latter, Penzias and Wilson used a large antenna to detect radio waves that were relayed by bouncing off of satellites. After they had removed interference effects due to radar, radio, and heat, they found residual ambient noise that couldn't be eliminated \cite{wiki:cosmic-radiation}. % \inlineitem{Both $T$ and $T^\star$ may be multiple, causing an individual process to fork into communicating sub-processes that involve different skills sets.} %% \item \textbf{Multiple influences}: The ``bridge'' from trigger to result is often found through a social network, thus, for instance Penzias and Wilson only understood the significance of their work after reading a preprint by Jim Peebles that hypothesised the possibility of measuring radiation released by the big bang \cite{wiki:cosmic-radiation}. % \inlineitem{The process as a whole may be multiplied out among different communicating investigators.} \end{itemize}

Materials, like gold, and processes, like metalurgy, have no value without a context of application: decoration, trade, circuitry, and so on. In practice, we are likely to attribute \emph{value} to materials that are useful, and \emph{creativity} to a person who puts materials to use in a novel way.

%% MC> Angelina is a able to read Twitter to find out what people think of %% MC> people like Hamid Karzai, and then change the sorts of images that %% MC> it finds as a result. So you're going to see a happy picture of %% MC> President Obama later next to a very angry picture of Hamid Karzai. %% MC> While some of this might look creative and intelligent, a lot of it %% MC> comes down to serendipity as well. So the image you're about to see %% MC> comes up for a Google search for terrorism that doesn't really have %% MC> much relevance to the news article, and the sound that you're %% MC> hearing now, the electronic drone, sounds like it's a good choice %% MC> for a game that's about war and about feeling unsettling. But in %% MC> actuality I have no idea how Angelina came up with that choice.

The primary source of serendipity triggers would be presentations or feedback that independently prepared systems find meaningful and useful. A typical example might be a poem shared by one system that another system finds particularly interesting. The listener might make a note to the effect ``I would like to be able to write like that'' or ``I hope that my poetry doesn't sound like that.'' In a typical Writers Workshop, used as intended, feedback might arrive that would cause the presenting system to change its writing. A more unexpected result would be for a system to change its \emph{genre}, e.g. to switch from writing poems to writing programs. Here's what might happen in a discussion of the first few lines of ``On Being Malevolent,'' written by an early user-defined flow chart in the {\sf FloWr} system (known at the time as {\sf Flow}) \cite{colton-flowcharting}. Note that for this dialogue to be possible, it would presumably have to be conducted within a lightweight process language, as discussed above. Nevertheless, for convenience, the discussion will be presented here as if it was conducted in natural language. Whether contemporary systems have adequate natural language understanding to have interesting interactions is one of the key unanswered questions of this approach, but protocols like the ones described above would be sufficient to make the experiment. \begin{center} \begin{minipage}{.9\textwidth} \begin{dialogue} \speak{Flow} ``\emph{I hear the souls of the damned waiting in hell. / I feel a malevolent spectre hovering just behind me / It must be his birthday}.'' % \speak{System A} I think the third line detracts from the spooky effect, I don't see why it's included. % \speak{System B} It's meant to be humourous -- in fact it reminds me of the poem you presented yesterday. % \speak{Moderator} Let's discuss one poem at a time. \end{dialogue} \end{minipage} \end{center} To the extent possible, exchanges in the process language should be a matter of dynamics rather than representation: this is another way to say that ``triggers'' should be independent of their ``results.'' Someone saying something in the workshop does not cause the participant to act, but rather, to think. % For example, even if, perhaps and especially because, cross-talk about different poems is bending the rules, the dialogue above could prompt a range of reflections and reactions. System A may object that it had a fair point that has not been given sufficient attention, while System B may wonder how to communicate the idea it came up with without making reference to another poem.

\subsection{Proposed experiment: A Writers Workshop for Systems} \label{sec:writers-workshop} Richard Gabriel \cite{gabriel2002writer} describes the practise of Writers Workshops that has been put to use for over a decade within the Pattern Languages of Programming (PLoP) community. The basic style of collaboration originated much earlier with groups of literary authors who engage in peer-group critique. Some literary workshops are open as to genre, and happy to accommodate beginners, like the Minneapolis Writers Workshop\footnote{\url{http://mnwriters.org/how-the-game-works/}}; others are focused on professionals working within a specific genre, like the Milford Writers Workshop\footnote{\url{http://www.milfordsf.co.uk/about.htm}}. The practices that Gabriel describes are fairly typical. Authors come with work ready to present, and read a short sample, which is then discussed and constructively critiqued by attendees. Presenting authors are not permitted to rebut these comments. The commentators generally summarise the work and say what they have gotten out of it, discuss what worked well in the piece, and talk about how it could be improved. The author listens and may take notes; at the end, he or she can then ask questions for clarification. Generally, non-authors are either not permitted to attend, or are asked to stay silent through the workshop, and perhaps sit separately from the participating authors/reviewers. There are similarities between the Writers Workshops and classical practices of group composition \cite{jin1975art} and dialectic \cite{dialectique}, and the workshop may be considered an artistic or creative space in its own right. In PLoP workshops, authors present design patterns and pattern languages, or papers about patterns, rather than more traditional literary forms like poems, stories, or chapters from novels. Papers must be workshopped at a PLoP or EuroPLoP conference in order to be considered for the \emph{Transactions on Pattern Languages of Programming} journal. A discussion of writers workshops in the language of design patterns is presented by Coplien and Woolf \cite{coplien1997pattern}. Their patterns include: \begin{center} {\small \begin{tabular}{l@{\hspace{.2cm}}l@{\hspace{.2cm}}l} \emph{Open Review} & \emph{Safe Setting} & \emph{Workshop Comprises Authors} \\ \emph{Authors are Experts} & \emph{Community of Trust} & \emph{Moderator Guides the Workshop} \\ \emph{Thank the Author} & \emph{Selective Changes} & \emph{Clearing the Palate} \\ \end{tabular} } \end{center} We propose that a similar pattern-based approach should be deployed within the Computational Creativity community to design a workshop in which the participants are computer systems instead of human authors. The annual International Conference on Computational Creativity (ICCC), now entering its sixth year, could be a suitable venue. Rather than the system's creator presenting the system in a traditional slideshow and discussion, or a system ``Show and Tell,'' the systems would be brought to the workshop and would present their own work to an audience of other systems, in a Writers Workshop format. This might be accompanied by a short paper for the conference proceedings written by the system's designer describing the system's current capabilities and goals. Subsequent publications might include traces of interactions in the Workshop, commentary from the system on other systems, and offline reflections on what the system might change about its own work based on the feedback it receives. As in the PLoP community, it could become standard to incorporate this sort of workshop into the process of peer reviewing journal articles for the new \emph{Journal of Computational Creativity}\footnote{\url{http://www.journalofcomputationalcreativity.cc}}. \begin{table}[p] \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Successful error}} & \\ \emph{Van Andel's example}: & Post-it\texttrademark\ notes \\[.2cm] {\tt presentation}& Systems should be prepared to share interesting ideas even if they don't know directly how they will be useful. \\ {\tt listening} & Systems should listen with interest, too. \\ {\tt feedback} & Even interesting ideas may not be ``marketable.''\\ {\tt questions} & How is your suggestion useful? \\ {\tt reflections} & New combinations of ideas take a long time to realise, and many different ideas may need to be combined in order to come up with something useful.\\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Side effect}} & \\ \emph{Van Andel's example}: & Nicotinamide used to treat side-effects of radiation therapy proves efficacious against tuberculosis. \\[.2cm] {\tt presentation}& Systems should use their presentation as an experiment. \\ {\tt listening} & Listeners should allow themselves to be affected by what they are hearing. \\ {\tt feedback} & Feedback should convey the nature of the effect.\\ {\tt questions} & The presenter may need to ask follow-up questions to gain insight. \\ {\tt reflections} & Form a new hypothesis before seeking a new audience. \\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Wrong hypothesis}} & \\ \emph{Van Andel's example}: & Lithium, used in a control study, had an unexpected calming effect. \\[.2cm] {\tt presentation}& How is this presentation interpretable as a (``natural'') control study? \\ {\tt listening} & Listeners are ``guinea pigs''.\\ {\tt feedback} & Discuss side-effects that do not necessarily correspond to the author's perceived intent. \\ {\tt questions} & Zero in on the most interesting part of the conversation.\\ {\tt reflections} & Revise hypotheses to correspond to the most surprising feedback. \\ \end{tabular} \bigskip \begin{tabular}{lp{.7\textwidth}} {\bf\emph{Outsider}} & \\ \emph{Van Andel's example}: & A mother suggests a new hypothesis to a doctor. \\[.2cm] {\tt presentation}& The presenter is here to learn from the audience. \\ {\tt listening} & The audience is here to give help, but also to get help.\\ {\tt feedback} & Feedback will inevitably draw on previous experiences and ideas.\\ {\tt questions} & What is the basis for that remark?\\ {\tt reflections} & How can I implement the suggestions?\\ \end{tabular} \vspace{.2cm} \caption{Reinterpreting patterns of serendipity for use in a computational workshop\label{tab:reinterpret}} \end{table} \begin{figure}[t] \begin{center} \resizebox{.93\textwidth}{!}{ \StickyNote[2.5cm]{myyellow}{{\LARGE {Interesting idea}} \\[4ex] {Surprise birthday party}}[3.8cm] \StickyNote[2.5cm]{mygreen}{{\Large I heard you say:} \\[4ex] {``surprise''} }[3.8cm] \StickyNote[2.5cm]{pink}{{\Large Feedback:} \\[4ex] {I don't like surprises}}[3.8cm] } \resizebox{.61\textwidth}{!}{ \StickyNote[2.5cm]{myorange}{{\LARGE {Question}} \\[4ex] {Not even a little bit?\ldots}}[3.8cm] \quad \raisebox{-.2cm}{\StickyNote[2.5cm]{myblue}{{\LARGE Note to self:} \\[4ex] {(Try smaller surprises \\ next time.)}}[3.8cm]} } \end{center} \caption{A paper prototype for applying the \emph{Successful Error} pattern\label{fig:paper-prototype}} \end{figure} In order to facilitate this sort of interaction, it would be necessary for systems to implement a basic protocol related to %% \[ \text{ {\tt presentation}, {\tt listening}, {\tt feedback}, {\tt questions}, and {\tt reflections}.} \] %% This protocol could be thought of as a light-weight template for creating design patterns that guide system-level participation in the context specified by Coplien and Woolf's pattern language for writers workshops. Table \ref{tab:reinterpret} uses this framework to recast the four ``perfectly'' serendipitous patterns from van Andel -- \emph{Successful error}, \emph{Side effect}, \emph{Wrong hypothesis}, and \emph{Outsider} -- in a form that may make them useful to developers preparing to enter their systems into the Workshop. % Further guidelines for structuring and participating in traditional writers workshops are presented by Linda Elkin in \cite[pp. 201-203]{gabriel2002writer}. It is not at all clear that the same ground rules should apply to computer systems. For example, one of Elkin's rules is that ``Quips, jokes, or sarcastic comments, even if kindly meant, are inappropriate.'' Rather than forbidding humour, it may be better for individual comments to be rated as helpful or non-helpful. Again, since serendipitous discovery is an overarching goal, in the first instance, usefulness and interest might be judged in terms of the criteria described in Section \ref{sec:evaluation-criteria}. We would need a neutral environment that is not hard to develop for: the {\sf FloWr} system described in Section \ref{sec:foundations} offers one such possibility. With this system, the basic operating logic of the Workshop could be spelled out as a flowchart, and contributing systems could use flowcharts as the basic medium for sharing their presentations, feedback, and questions. Developing around a process language of this sort partially obviates the need for participating systems to have strong natural language processing capabilities. % Post-it\texttrademark\ notes, which have provided us with a useful example of serendipitous discovery, also provide indicative strategies from the world of paper prototyping (Figure \ref{fig:paper-prototype}). Gordon Pask's conversation theory, reviewed in \cite{conversation-theory-review,boyd2004conversation}, goes considerably beyond what we have presented here as a simple process language, although there are structural parallels. In a basic Pask-style learning conversation: (0) Conversational participants are carrying out some actions and observations; (1) naming and recording what action is being done; (2) asking and explaining why it works the way it does; (3) carrying out higher-order methodological discussion; and (4) trying to figure out why unexpected results occured \cite[p. 190]{boyd2004conversation}. Naturally, variations to the underlying system, protocol, and the schedule of events should be considered depending on the needs and interests of participants, and several variants can be tried. On a pragmatic basis, if the Workshop proved quite useful to participants, it could be revised to run monthly, weekly, or continuously.\footnote{For a comparison case in computer Go, see \url{http://cgos.computergo.org/}.}

%% \textbf{[It would be good to go back over our other paper and make %% sure we make good on the idea in the Related Work section of the %% current paper that ``This earlier paper remains broadly %% indicative, however, and the ideas it describes can see %% considerable benefit from the more formal thinking we develop in %% the current work.''} %% \textbf{In particular: at least one of the reviewers found the Writers %% Workshop ``technologically unrealistic'' or similar, so let's try to %% make sure we're not overpromising. I think the other paper makes it %% all fairly realistic.]} %% In \cite{poetry-workshop}, we investigate the feasibility of using %% designs of this sort in multi-agent systems that learn by sharing and %% discussing partial understandings and making social revisions. This earlier paper remains broadly %% indicative, however, and the ideas it describes can see considerable %% benefit from the more formal thinking we develop in the current work. % \citeA{poetry-workshop} describes a Writers Workshop for poetry %systems.

In the next subsection we will review several historical examples. First, one further point should be made with reference to the ``The Three Princes of Serendip''. Prior to Walpole's coinage, this story had been adapted by Voltaire into an early chapter of \emph{Zadig}, and in turn ``the method of Zadig'' informed subsequent approaches both to fiction writing and natural science. This method is rooted firstly in discovery: \begin{quote} ``[Zadig] \emph{pry’d into the Nature and Properties of Animals and Plants, and soon, by his strict and repeated Enquiries, he was capable of discerning a Thousand Variations in visible Objects, that others, less curious, imagin’d were all alike.}''~\cite[pp. 21--22]{zadig} \end{quote} \noindent Secondly, from disparate observations, Zadig is often able to assemble a coherent picture: \begin{quote} \emph{It was his peculiar Talent to render Truth as obvious as possible: Whereas most Men study to render it intricate and obscure.}~\cite[p. 54]{zadig} \end{quote}