Authorea

Joe Corneli fix some bibliography entries and a footnote about 9 years ago

Commit id: ad2a7a8f39e7c655dfb801eda74b601cfb0e26af

deletions | additions

Finally, we ask, again, comparing similar results where possible: \emph{($\mathbf{d}$: \textbf{value})} How valuable is the result that is ultimately produced?} \medskip % %Then combining $\mathbf{a}\times\mathbf{b}\times\mathbf{c}$ gives a % likelihood score: \emph{Low likelihood $\mathbf{a}\times\mathbf{b}\times\mathbf{c}$

\begin{quote} {\em Test your serendipitous system against the standards stated in Step 2 and report the results.}\end{quote} \noindent In Section \ref{sec:computational-serendipity} we pilot our framework by examining the degree of serendipity of existing computational systems, systems and looking for ways that they their serendipity could become more serendipitous be enhanced. We will also use the framework to guide the high-level design of a novel system.

% Most prior work that deals with serendipity in a computing context focuses on computational ``discovery''; we argue that serendipity also includes an important ``invention'' aspect. % We survey literature describing serendipitous discovery and invention in science and technology, as well as the etymology and definitions of the term ``serendipity''. We propose a model of computational serendipity, building Building upon and refining previous work. This model suggests work, we propose a definition model of computational serendipity that can be used to evaluate computational systems. To this end we adapt existing recommendations for evaluating computational creativity, to fully develop and apply our model for evaluation of computational serendipity. \emph{creativity}. % We develop case studies that evaluate the serendipity of existing systems, and develop a thought experiment that applies our model toa design for a multi-agent environment for computer poetry. % From these our analyses, we extract recommendations for practitioners working with computational serendipity, and outline future directions. directions for research. \\[.5cm] % %% \keywords{serendipity,

title={Chase, chance, and creativity: The lucky art of novelty}, author={Austin, James H}, year={2003 [1978]}, publisher={Mit publisher={MIT Press} } @article{mccrae1987creativity,

@article{cropley2006praise, title={In praise of convergent thinking}, author={Cropley, Arthur}, journal={Creativity research journal}, Research Journal}, volume={18}, number={3}, pages={391--404},

@book{bergson1983creative, Author = {Bergson, Henri}, Publisher = {Henry Holt and Company}, Title = {Creative evolution}, {E}volution}, Year = {1911 [1907]}} @article{milan2013kiki,

@inproceedings{McNee2006, Author = {McNee, SM and Riedl, J and Konstan, JA}, Booktitle = {CHI'06 {{CHI}'06 extended abstracts on {H}uman factors in computing systems}, File = {:Users/worldwindow/Documents/Mendeley Desktop/McNee, Riedl, Konstan - 2006 - Being accurate is not enough how accuracy metrics have hurt recommender systems.pdf:pdf}, Isbn = {1595932984}, Keywords = {collaborative,personalization,recommender systems},

Keywords = {accuracy,collaborative filtering,diversification,metrics,novelty,recommender systems,serendipity}, Mendeley-Groups = {Magisterarbeit/Recommendation and Serendipity}, Pages = {13--22}, Title = {{Auralist : Introducing Serendipity {Auralist: {I}ntroducing {S}erendipity into Music Recommendation}}, {M}usic {R}ecommendation}, Year = {2011}} @inproceedings{Sugiyama2011,

File = {:Users/worldwindow/Documents/Mendeley Desktop/Onuma, Tong, Faloutsos - 2009 - TANGENT A Novel 'Surprise-me' Recommendation Algorithm.pdf:pdf}, Isbn = {9781605584959}, Mendeley-Groups = {Magisterarbeit/Recommendation and Serendipity}, Title = {{TANGENT : A Novel 'Surprise-me' Recommendation Algorithm}}, {{TANGENT}: {A} {N}ovel {`}{S}urprise-me{'} {R}ecommendation {A}lgorithm}, Url = {http://dl.acm.org/citation.cfm?id=1557093}, Year = {2009}, Bdsk-Url-1 = {http://dl.acm.org/citation.cfm?id=1557093}}

\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.'' It is important to notice that \emph{he} was in the right spot at the right time as well -- and that this was not a complete coincidence. The chance events we're interested in always include at least one observer. \end{itemize} \begin{itemize}

begin or to continue a search into unfamiliar territory. We use this word to describe both simple curiousity and related deeper drives. Charles Goodyear \citeyear{goodyear1855gum} reflects on his own life experience: experience as follows: ``from the time his attention was first given to the subject, a strong and abiding impression was made upon his mind, that an object so desirable and important, and so necessary to man's comfort, as the making of gum-elastic available to his use,

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.'' One quite literal examplethat takes this literally is the Swiss company Freitag, started by design students who built a business around ``upcycling'' used truck tarpaulins into bags and backpacks. Thanks in part to clever marketing \cite[pp. 54--55,

\citeyear[p. 643]{van1994anatomy} estimates that in twenty percent of innovations ``something was discovered before there was a demand for it.'' To illustrate the role of this factor, it may be most revealing to consider a \emph{counterexample}, ``counterexample,'' in which dynamics are were not attended to carefully and the process suffers as a result. Cropley \citeyear{cropley2006praise} describes Eugen Semmer's failure to recognise the role of \emph{penicillium notatum} in restoring two unwell horses to health: ``Semmer saw the horses' return to good health as a problem that made it impossible for him to investigate the cause of their death, and reported [$\ldots$] \ldots\ on how he had succeeded in eliminating the mould from his laboratory!'' \end{itemize}

\noindent We will show how the key condition, the components, dimensions and environmental factors of serendipity can be modelled and assessed in computational systems in Section Sections \ref{sec:our-model} and \ref{sec:computational-serendipity}.

presupposes a smart mind,'' and these examples suggest potential directions for further work in computational intelligence. We then turn to a more elaborated thought experiment that describes a new system designed design that has been created with our criteria serendipity in mind. Before describing these examples, as a baseline, we introduce the notion of \emph{minimally serendipitous systems}. According to our

\subsection{Case Studies: Prior art} \label{sec:priorart} \paragraph{Evolutionary \paragraph{An evolutionary music improvisation systems.} system.} \citeA{jordanous10} reported a computational jazz improvisation system using genetic algorithms. Genetic algorithms, and evolutionary computing more generally, could encourage computational serendipity. We examine Jordanous's system (later given the name {\sf GAmprovising} \cite{jordanous:12}) as a case study for evolutionary computing in the context of our model of computational serendipity: to what extent does {\sf GAmprovising} model serendipity?

The {\sf GAmprovising} system can be said to have a \textbf{prepared mind} through its background knowledge of what musical concepts to embed in the Improvisors and the evolutionary abilities to evolve Improvisors. A \textbf{serendipity trigger} comes from the combination of the mutation and crossover operations previously employed in the genetic algorithm, and the user input feeding into the fitness function to evaluate produced music. A \textbf{bridge} is built by through creation of new Improvisors. The \textbf{results} are the various musical improvisations produced by the fittest Improvisors (as well as, perhaps, the parameters that have been considered fittest). The likelihood of serendipitous evolution is greatly enhanced by the use of random mutation and crossover operations within the genetic algorithm, which increase the diversity of search space covered by the system during evolution. The \textbf{chance} of encountering any particular pair of Improvisor and user evaluation is vanishingly low, given the massive dimensions of the search space. The evolution of the population of Improvisors could be described as \textbf{curiosity} about how to satisfy the musical tastes of a particular human user who identifies certain Improvisors as interesting. The system's \textbf{sagacity} corresponds to the likelihood that the user will appreciate a given Improvisor's music (or similar music) over time. One challenge here is that the tastes of the user may change. The \textbf{value} of the results are maximised through employing a fitness function. Evolutionary systems such as {\sf GAmprovising} necessarily operate in a \textbf{dynamic world} which is evolving continuously and that must, in particular, take into account the evolution of the user's tastes. The \textbf{multiple \textbf{Multiple contexts} arise from the user changing their preferences over time or through the possibility of having multiple users evaluate the musical output or through the user changing their preferences over time. A output. This variant version of the systemthat would cater to multiple users is not yet implemented formally -- a revised system with these features implemented, but would be curious about occupied with the more complex problem of satisfying multiple different users' preferences simultaneously. Moving to a more complex problem domain would require the system to be curious about more than one user at a time, and require greater sagacity if the system is to successfully satisfy multiple tastes. \textbf{Multiple tasks} are carried out by the system including evolution of Improvisors, generation of music by individual Improvisors, capturing of user ratings of a sample of the Improvisors' output, and fitness calculations. \textbf{Multiple influences} are captured through the various combinations of parameters that could be set and the potential range of values for each parameter. %% Table \ref{caseStudies} summarizes how serendipity in such a system can be described in terms of our model. \paragraph{Recommender systems.} % Stress distinction between serendipity on the system- vs. serendipity on the user's side. As discussed in Section \ref{sec:related}, recommender systems are one of the primary contexts in computing where serendipity is considered. Most discussions of serendipity in recommender systems focus on suggesting items to a user that will be likely to introduce new ideas that are unexpected, but close to what the user is already interested in. A recommendation of this type will be called (possibly pseudo-)serendipitous. \mbox{pseudo-)serendipitous}. As we noted, these systems mostly focus on supporting discovery, but some architectures also seem to take account of invention, such as the Bayesian methods surveyed in Chapter 3 of \citeNP{shengbo-guo-thesis}. Recommender systems \emph{stimulate} serendipitous discovery, by \emph{simulating} when this is likely to occur. In respect to related work, we therefore have to distinguish serendipity on the the user side from serendipity in the system. As we have indicated, most current Current research in this area focuses on the first aspect and tries to find and assess \textbf{serendipity triggers} by exploiting patterns in the search space. For example, \citeA{Herlocker2004} as well as \citeA{Lu2012} associate less popular items with high unexpectedness. Clustering is also frequently used to discover latent structures in the search space. For example, \citeA{Kamahara2005} partition users into clusters of common interest, while \citeA{Onuma2009} as well as \citeA{Zhang2011} perform clustering on both users and items. In the work by \citeA{Oku2011}, %\citeA{Oku2011} allow the useris allowed to select two items in order to mix their features features. in a sort of conceptual blend. Note that in the course of evolution of these and other systems it is generally the system's developers who plan and perform adaptations; adaptations: even in the Bayesian case, the system has limited autonomy. Nevertheless, the impetus to develop increasingly autonomous recommender systems is present, especially in complex domains where hand-tuning is either very cost-intensive or infeasible. With this challenge in mind, we investigate how serendipity could be achieved on the system side, and potentially be reflected back to the user. In terms of our model, current systems have at least the makings of a \textbf{prepared mind}, comprising both a user- and a domain model, both of which can be updated dynamically. User behaviour (e.g.~following up on these recommendations) may serve as a \textbf{serendipity trigger} for the system, and change the way it makes recommendations in the future. A \textbf{bridge} to a new kind of recommendation may be found by pattern matching, and especially by looking for exceptional cases: when new elements are introduced into the domain which do not cluster well, or different clusters appear in the user model that do not have obvious connections between them. The intended outcome of recommendations depends on the organisational mission, and can in most cases be situated between making money and empowering the user. The serendipitous \textbf{result} on the system side would be learning a new approach that helps to address these goals. %%% \begin{table}[Ht!]

\cline{2-3} \textbf{Dynamic world} & Continuous computational evolution and changes in user tastes& As precondition for testing system's influences on user behaviour\\ %\cline{2-3} \textbf{Multiple contexts} & Multiple users users' opinions-- would change what the curiousity profile system is curious about and require greater sagacity & User model and domain model\\ % \cline{2-3} \textbf{Multiple tasks} & Evolving Improvisors, generating music, collecting user input, fitness calculations & Making recommendations, learning from users, updating models \\ % \cline{2-3}

\normalsize %%% The imperfect knowledge about the user's preferences and interests represents a main source of \textbf{chance}. Combined with the ability to learn, \textbf{curiosity} could be described as the urge to make recommendations specifically for the purposes of finding out more about users, possibly to the detriment of other metrics over the short term. Measures of \textbf{sagacity} would relate to the system's ability to draw useful inferences from user behaviour. For example, the system might decide to initiate an A/B test to decide how a novel recommendation strategy influences conversion. The \textbf{value} of recommendation strategies can be measured in terms of traditional business metrics or other organisational objectives. Recommender systems have to cope with a \textbf{dynamic world} of changing user preferences and a changing collection of items to recommend. A dynamic environment which nevertheless exhibits some degree of regularity represents a precondition for useful A/B testing. As mentioned above the primary \textbf{(multiple) contexts} are the user model and the domain model. A system matching the description here would have \textbf{multiple tasks}: making useful recommendations, generating new experiments to learn about users, and building new models. Such a system could avail itself of \textbf{multiple influences} related to experimental design, psychology, and domain understanding. \medskip Table \ref{caseStudies} summarises how the components, dimensions and factors of our model of serendipity can be mapped to evolutionary music systems and the class of ``next-generation'' recommender systems discussed above. These case studies have shown how our model can be used to highlight the aspects of existing systems that would need to be developed further to enhance measures of the system's serendipity. % As a general comment, we would say that this is largely how % \emph{research and development} of recommender systems works, but

% We began by surveying ``serendipity'', developing a broad historical view, and describing several criteriafor serendipity which we propose to be computationally salient. We reviewed related work; like \citeA{andre2009discovery}, we propose a two-part definition of serendipity: \emph{discovery} followed by \emph{invention}.

Systems'' (SPECS) model from \citeA{jordanous:12}, we developed a set of evaluation standards for serendipity. % We used this model to analyseprior examples of serendipity in the context of evolutionary music improvisation and recommender systems, and developed a thought experiment that seems able to support ``high serendipity'' with a novel design for describes a computational poetry workshop. system that can (sometimes) make ``highly serendipitous'' creative advances in computer poetry, without user intervention. % We then reflected back over our definition, outlining definition and analyses, and outlined a programme for serendipitous computing in the pursuit of \emph{autonomy}, \emph{learning}, \emph{sociality}, and \emph{embedded evaluation}. We posit the following challenges, which connect with ongoing discussions

% We indicate several possible further directions for implementation work in each of our case studies. We have also drawn attention to theoretical questions related to program design in an autonomous programming context. Our examples show that serendipity is not foreign to computing practice. There are further gains to be had for research in computing by planning -- and programming -- for

\section{Discussion} \label{sec:discussion} In the preceding section, we applied our model to develop evaluate the serendipity of an evolutionary music improvisation system and a class of next-generation recommender systems, and assess we sketched a computationally serendipitous system, design for a multi-agent system for poetry based on the idea of a writers workshop. Writers Workshop. The model has helped to highlight directions for development that would increase the potential for serendipity in existing systems, either incrementally or more transformatively. The model has helped create a reasonably concrete system design. Our analysis of these examples illustrates some of steps that can be taken in order to design systems that can observe events that would otherwise not be observed, take an interest in them, and transform their observations into an artefact with lasting value. We will now discuss implications from our findings for future research, and outline potential next steps. \input{recommendations} \input{future-work-intro}

%% Here we do not mean to suggest that every instance of ``a solution to a %% problem in a context'' is due to serendipity at work -- on the %% contrary, that is just the discovery step. Inventing a viable design pattern %% only happens when the solution is found to be explicable and useful. To van Andel's assertion that ``The very moment I can plan or programme `serendipity' it cannot be called serendipity anymore,'' we

serendipity can be rewritten as a design pattern using the template suggested by our model; in future work, we would aim to build a more complete pattern language along similar lines. % The example pattern describes a scenario that is quite close tothe Pease et al.'s \citeyear{pease2013discussion} description of an online system that gathers new modules over time, and for which, periodically, new combinations of modules can yield new and interesting results.

guidelines for human programmers and convey a sense of our long-term plans for serendipitous computing systems. \begin{figure}[t] \begin{mdframed} \paragraph{\textbf{Successful error}}~ \vskip -1\baselineskip \begin{flushright}\emph{Van Andel's example} -- Post-it\texttrademark\ Notes \end{flushright} \vspace{-.15cm} \begin{description}[itemsep=2pt] \item[{\tt context}] -- You run a creative organisation with several different divisions and many contributors with different expertise. \item[{\tt problem}] -- One of the members of your organisation discovers something with interesting properties, but no one knows how to turn it into a product with industrial or commercial application. \item[{\tt solution}] -- You create a space for sharing and discussing interesting ideas on an ongoing basis (perhaps a Writers Workshop). \item[{\tt rationale}] -- You suspect it's possible that one of the other members of the firm will come up with an idea about an application; you know that if a potential application is found, it may not be directly marketable, but at least there will be a prototype that can be concretely discussed. \item[{\tt resolution}] -- The \emph{Successful error} pattern rewritten using this template is an example of a similar prototype, showing that serendipity can be talked about in terms of design patterns. \end{description} \end{mdframed} \caption{Our design pattern template applied to van Andel's \emph{Successful error} pattern\label{fig:va-pattern-figure}} \end{figure}

%% \end{itemize} %% \citeA{meszaros1998pattern} describe the typical scenario for authors of design patterns: \begin{figure}[!ht] \begin{mdframed} \paragraph{\textbf{Successful error}}~ \vskip -1\baselineskip \begin{flushright}\emph{Van Andel's example} -- Post-it\texttrademark\ Notes \end{flushright} \vspace{-.15cm} \begin{description}[itemsep=2pt] \item[{\tt context}] -- You run a creative organisation with several different divisions and many contributors with different expertise. \item[{\tt problem}] -- One of the members of your organisation discovers something with interesting properties, but no one knows how to turn it into a product with industrial or commercial application. \item[{\tt solution}] -- You create a space for sharing and discussing interesting ideas on an ongoing basis (perhaps a Writers Workshop). \item[{\tt rationale}] -- You suspect it's possible that one of the other members of the firm will come up with an idea about an application; you know that if a potential application is found, it may not be directly marketable, but at least there will be a prototype that can be concretely discussed. \item[{\tt resolution}] -- The \emph{Successful error} pattern rewritten using this template is an example of a similar prototype, showing that serendipity can be talked about in terms of design patterns. \end{description} \end{mdframed} \caption{Our design pattern template applied to van Andel's \emph{Successful error} pattern\label{fig:va-pattern-figure}} \end{figure} \noindent ``You are an experienced practitioner in your field. You have noticed that you keep using a certain solution to a commonly occurring problem. You would like to share your experience with others.'' There are many ways to describe a solution.

An interdisciplinary perspective on the phenomenon of serendipity promises further illumination. Here, we consider the potential for formalising this concept. This paper follows and expands \citeA{pease2013discussion}, where many of the ideas that are developed here were first presented. The current paper reassesses and updates this earlier work, developing it towards a computational characterisation of serendipity for computer modelling and system evaluation. New claims are advanced, positioning serendipity as a fundamental concept in computational creativity, with exciting potential to play a key role in computational intelligence more broadly. There is particularly interesting potential for serendipity within computational systems whose processes involve interaction with users.\footnote{It users, and autonomous systems that make use of a multi-agent framework.\footnote{It should not be assumed that a system that can accommodate user interaction would directly lead to serendipity; take for example the use of a calculator, where potential for serendipity through user interaction is minimal at best.} Serendipity isitself centred on reevaluation. For example, a non-sticky ``superglue'' that no one was quite sure how to use turned out to be just the right ingredient for 3M's Post-it\texttrademark\ notes.

\noindent This definition can be summarised schematically as follows, with letters referencing to the key condition and components introduced in the literature survey: % \input{schematic-tikz} {\centering \includegraphics[width=.8\textwidth]{schematic} \input{schematic-tikz} %\includegraphics[width=.8\textwidth]{schematic} \par}

\begin{tikzpicture}[auto, node distance=2cm,>=latex'] \node [sum] (sum1) {}; \node [input, name=pinput, above left=.7cm left=.9cm and .7cm .9cm of sum1] (pinput) {}; \node [input, name=tinput, left=2cm of sum1] (tinput) {}; \node [input, name=minput, below left of=sum1] (minput) {}; \node [input, name=minput, right of=sum1] (moutput) {}; \draw [->] (tinput) -- node{\vphantom{{\tiny g}}{\tiny context}} node{\vphantom{{\footnotesize g}}{\footnotesize context~}} (sum1); \draw [->] (pinput) -- node{{\tiny node{{\footnotesize problem}} (sum1); \draw [->] (sum1) -- node{\vphantom{{\tiny g}}{\tiny node{\vphantom{{\footnotesize g}}{\footnotesize solution}} (moutput); \end{tikzpicture} \hspace{1cm} \begin{tikzpicture}[auto, node distance=2cm,>=latex'] \node [sum] (sum1) {}; \node [input, name=pinput, above left=.7cm left=.9cm and .7cm .9cm of sum1] (pinput) {}; \node [input, name=tinput, left of=sum1] (tinput) {}; \node [input, name=minput, below left of=sum1] (minput) {}; \node [sum, right=1.5cm of sum1] (sum2) {}; \node [input, name=minput, right of=sum2] (moutput) {}; \draw [->] (tinput) -- node{\vphantom{{\tiny g}}{\tiny node{\vphantom{{\footnotesize g}}{\footnotesize solution}} (sum1); \draw [->] (pinput) -- node{{\tiny node{{\footnotesize rationale}} (sum1); \draw [->] (sum1) -- node{\vphantom{{\tiny g}}{\tiny node{\vphantom{{\footnotesize g}}{\footnotesize pattern}} (sum2); \draw [->] (sum2) -- node[text width=1.5cm,execute at begin node=\setlength{\baselineskip}{.3ex}]{\tiny node=\setlength{\baselineskip}{.9ex}]{\footnotesize \emph{resolution\\~of forces}} (moutput); \end{tikzpicture} \endgroup \end{center}

challenges for research in computational serendipity. \begin{itemize} \item \textbf{Autonomy}: Our case study on serendipity in recommender systems studies in Section \ref{sec:priorart} highlights highlight the need for more potential value of increased autonomy on the system side. %% The thought experiment in Section %% \ref{sec:ww} develops a design illustrating the relationship between %% creativity at the level of artefacts (e.g.~new poems) and %% creativity at the level of \emph{problem specification} (learning %% new poetic concepts). The search for connections that make raw data into ``strategic data'' is an appropriate theme for research in computational intelligence and machine learning to grapple with. In the standard cybernetic model, we control computers, and we also control the computer's operating context. There is little room for serendipity if there is nothing outside of our direct control. In contrast with the mainstream model, von Foerster \citeyear[p. 286]{von2003cybernetics} advocated a \emph{second-order cybernetics} in which ``the observer who enters the system shall be allowed to stipulate his own purpose.'' \emph{A primary challenge to the serendipitous operation of computers is developing computational agents that specify their own problems.} \end{itemize} \begin{itemize} \item \textbf{Learning}: The Writers Workshop described in Section \ref{sec:ww} is fundamentally a one possible designsketch for a system that can \emph{learn from experience}. The Workshop model ``personifies'' the wider world in the form of one or several critics. It is clearly also possible for a lone creative agent to

\end{itemize} \begin{itemize} \item \textbf{Sociality}: We may be aided in our pursuit of the ``smart mind'' required for serendipity by recalling Turing's proposal that computers should ``be able to converse with each other to sharpen their wits'' \cite{turing-intelligent}. Other fields, including computer Chess, Go, and argumentation have achieved this, and to good effect.Deleuze \citeyear[p. 26]{deleuze1994difference} wrote: ``We learn nothing from those who say: `Do as I do'. Our only teachers are those who tell us to `do with me'[.]'' Turing recognised that computers would have to be coached in the direction of social learning, but that once they attain that standard they will learn much more quickly. Deleuze \citeyear[p. 26]{deleuze1994difference} wrote: ``We learn nothing from those who say: `Do as I do'. Our only teachers are those who tell us to `do with me'[.]'' \emph{A third challenge is for computational agents to interact in a recognisably social way with us and with each other, resulting in emergent effects.} \end{itemize} \begin{itemize}

\citeA{stakeholder-groups-bookchapter} outlined a general programme for computational creativity, and examined perceptions of creativity in computational systems found among members of the general public, Computational Creativity researchers, and existing creative (human )communities. communities. We should now add a fourth important ``stakeholder'' group in computational creativity research: computer systems themselves. Creativity may look very different to this fourth stakeholder group than it looks to us. We should help computers evaluate their own results It is our responsibility as system designers to teach our systems how to make evaluations in way that is both reasonable and creative process. ethical. This is exemplified by the preference for a ``non-zero sum'' criterion for value suggested in our discussion of the dimensions of serendipity in Section \ref{sec:by-example}. \emph{A fourth challenge is for computational agents to evaluate their own creative process and products.} \end{itemize} It is our responsibility as system designers to teach them how to make evaluations in way that is both reasonable and ethical. This is exemplified by the preference for a ``non-zero sum'' criterion for value suggested in our discussion of the dimensions of serendipity in Section \ref{sec:by-example}. A survey of word occurrences from a recent special issue of \emph{Cognitive Computation} on ``Computational Creativity, Intelligence and Autonomy'' \cite{bishop-erden-special-issue} shows that related themes are broadly active in the research community.\footnote{Articles community. Here \emph{italics} indicates that the word stem accounted for 0.1\% of the article or more; added \textbf{\emph{bold}} indicates that it accounted for 1\% or more.\footnote{Articles were converted to text via {\tt pdftotext -layout}, individual counts found via {\tt tr \textquotesingle~\textquotesingle~\textquotesingle\textbackslash n\textquotesingle~< file.txt | grep -c "stem*"}, and total word counts via {\tt wc -w}. The corresponding counts for the \emph{current} paper are 10, \emph{27}, 12, \emph{25}, \emph{16}, \emph{45} \emph{46} and 12.8K.} Here \emph{italics} indicates that the word stem accounted for 0.1\% of the article or more; added \textbf{\emph{bold}} indicates that it accounted for 1\% or more. 12.5K.} \medskip

\begin{center} \begingroup \tikzset{ block/.style = {draw, fill=white, rectangle, minimum height=3em, minimum width=3em},

\draw [->] (sum2) -- node{$|R|>0$} (moutput); \end{tikzpicture} \endgroup \end{center}

\newcommand{\sunmark}{\ding{106}}% \newcommand{\handmark}{\ding{43}}% % \usepackage{lineno} % \usepackage{pagecolor} % \pagecolor{yellow!10!orange!5} \usepackage[framemethod=tikz]{mdframed} \mdfsetup{ skipabove=\baselineskip, skipbelow=0\baselineskip, innertopmargin=3pt, innerbottommargin=3pt innerbottommargin=3pt, apptotikzsetting={\tikzset{mdfbackground/.append style={fill=red,fill opacity=0}}} } %% \usepackage{fontspec} %% \newfontfamily{\tam}[Script=Tamil]{Lohit Tamil} %% \defaultfontfeatures{Scale=MatchLowercase}

\maketitle % \linenumbers \input{abstract.tex} \setcounter{footnote}{0}

In our thought experiment, we focus on the case of hypothetical discussions and exchange of views between computational poetry systems as our example of a situation where social circumstances could encourage serendipity. We note that similar ideas would apply for prose and, with further adaptation, other arts.\\ \noindent \textbf{Prepared mind.} Participating systems need to be able to follow the Workshop protocol.This means that participating systems will need components like those listed above. The {\tt listening} and {\tt questions} components stages of the protocol correspond to $p$ and $p^{\prime}$ our model of serendipity. The corresponding ``comment generator'' and ``feedback integrator'' modules in the architecture architectural sketch represent the primary points of interface to the outside world. between author and critic. In principle these modules need to be prepared to deal (more deal, more or less thoughtfully) thoughtfully, with \emph{any} text, and in turn, with \emph{any} comment on that text. Certain limits may be agreed in advance; e.g.~as to genre or length in the case of texts, and what constitutes an acceptable comment.The ``feedback explainer'' is closely connected with the ``comment generator'' and in an implementation of this model they would presumably share a codebase. %% The loop for learning by asking questions as they arise is %% reminiscent of the operating strategy of {\sf SHRDLU} %% \cite{winograd1972understanding}. Importantly, one of the most relevant preparations would be prior participation in Workshop dialogues. A participating system -- particularly one with prior experience in the Workshop may -- will have a catalogue of outstanding unresolved, or partially resolved, problems (denoted ``X'' in the schematic above). Figure \ref{fig:generative-diagram}). Embodied in code, they may these drive comments, questions, and other behaviour -- and they may be answered addressed in unexpected ways.\\ ways.\par\medskip \noindent \textbf{Serendipity triggers.} Although the poem is under the control of the initial generative

The listening subsystem expects some poem, but it does not know what poem to expect. In this sense, the poem constitutes a serendipity trigger $T$, not only for the listening subsystem, but for the Workshopsystem as a whole. % To expand this point, note that there may be several listeners, each sharing their own feedback and listening to the feedback presented by others (which, again, is outside of their direct control). This creates further potential for serendipity, since each listener can learn what others see in the poem. More formally, in this case $T^\star$ may be seen as an evolving vector with shared state, but viewed and handled from different perspectives. With multiple agents involved in the discussion, the ``comment generator'' module would expand to contain its own feedback loops.\\ loops.\par\medskip \noindent \textbf{Bridge.} Feedback on portions of the poem may lead the system to identify new

%% the computer \cite{winograd1972understanding}. However, it would %% be hard to call that ``serendipity.'' This sort of system extension is quite typical when a human programmer is involved. Hoever, However, here we are interested in the possibility of agents building new poetic concepts \emph{without} outside intervention, starting with some basic concepts and abilities related to poetry (e.g.~definitions of words, valence of sentiments, metre,

effective to encourage incidents onto which we might project the word serendipity.'' One cognitively inspired hypothesisthat could describe the serendipitous formation of new concepts isthe notion that the development of new concepts is closely related to formation development of new sensory experiences \cite{milan2013kiki}. %% If the workshop %% participants have the capacity to identify the distinctive features of %% a given poem, then training via a machine learning or genetic %% algorithm approach could be used to assemble a battery of existing %% low-level tools that can approximate the effect. Relatedly, a %% compression process could seek to produce a given complex poetic %% effect with a maximally-succinct %% algorithm \cite{schmidhuber2007simple}. % The key point is that feedback Feedback on the poem -- simply describing what is in the poem from several different points of view -- can be used to define new problems for the system to solve. %% This is not simply a %% matter of decomposing the poem into pieces, but also of reconstructing %% the way in which the pieces work together. This is one One of the functions of the {\tt questions} step step, corresponding to $p^{\prime}$ in our formalism: they offer formalism, is to give the poet the opportunity to enquire about how different pieces of feedback fit together, and learn more about where they come from. Although computers are currently nowhere close, the The reconstructive process may steadily approach the ideal case -- familiar to humans -- of relating to the sentiment expressed by the poem as a whole \cite[p. 209]{bergson1983creative}.\\ 209]{bergson1983creative}.\par\medskip %% Several of us are involved with a contemporary project %% \cite{coinvent14} to develop a formal theory of concept invention, %% focusing on \emph{concept blending}. The additive or subtractive %% blending of existing poetry profiles may be another way to create new %% concepts. %% should be possible Modifer Grammar %% Counting Breathing Position Distribution Phonics Rhythm Repetition %% Thematic Narrative Entropy \noindent \textbf{Result.} The final step is to take the problem or problems that were identified, and write new code to solve them. Several strategies for generating a result $R$, in the form of new code, were described above. Now the system evaluates the new code to see whether it holds promise. In order to do this, it must have a way to carry out an evaluation and judge whether $|R|>0$. In the most straightforward case, it the poet would simply make changes to the draft poem that seem to improve it in some way. For example, the poet might remove or alter material that elicited a negative response from a critic. The system may then proceed to update its modules related to poetry generation. Notably, it It may also update its own feedback modules, after reflecting on questions like: ``How might the critic have detected noticed that feature in my poem?'' poem?''\par\medskip \paragraph{Likelihood \noindent \textbf{Likelihood scores and potential value.} We can assume that most Assuming the poems consumed by presented to the system would never have been seen before, and are not too repetitive, the chance of observing encountering a given serendipity trigger would bevery small. While the likelihood of detecting an \emph{already-known} poetic feature (e.g.~rhyming, alliteration) in It should be straightforward for a given poem where it is present critic to detect some known feature, like metre or rhyme, but at least moderately difficult to notice a novel poetic idea. There is presumably fairly high, we are particularly interested in some nuance here, since whenever the cases of where system learns a \emph{novel} poetic feature can be discovered, new concept, the low-hanging fruit from the pool of new concepts is used up, and this would remain a relatively infrequent occurrence. the system's perceptiveness simultaneously increases. The chance that a newly-observed feature will result in usable code seems relatively high, although this remains to be shown in practice. Presumably many novel poetic ideas would be seen as high-value at first, while but only some of these new ideas will prove to have lasting value. Our likelihood score would be $\mathit{low}\times\mathit{low}\times\mathit{high}$, $\mathit{low}\times\mathit{medium}\times\mathit{high}$, or fairly low overall, and value would be varied, with at least some high-valued cases deserving meriting the description ``highly serendipitous.''\\ serendipitous.''\par\medskip \paragraph{Environmental \noindent \textbf{Environmental factors.} The system would set up its own internal dynamics, but it could also provide an interface for human poets to share their poetry andtheir critical remarks. There is one primary context, the Workshop, shared by all participants. The primary tasks envisaged in the system design are \emph{poetry generation} generation}, \emph{comment generation}, and \emph{code generation}. Although these are different tasks, they may have similar features (i.e.~both (i.e.~they all may present opportunities to learn from feedback). Influences could be highly multiple, including many very different kinds of poetry and various approaches from NLP.

%% {\centering %% \includegraphics[width=.9\textwidth]{ww-serendipity-diagram} %% \par} Italicised The italicised elements (\emph{presentation}, \emph{questions}, and \emph{reflections}) are the responsibilities of the presenting author; the other elements (listening, feedback, and answers) are the responsibilities of the attendant critics. % \begin{figure} \begin{figure}[b] {\centering \input{ww-generative-tikz}

\end{figure} % The system as a whole can be further decomposed into generative components, as in Figure \ref{fig:generative-diagram}. In our thought experiment, we focus on exploring the features of serendipity by thinking through how this design would apply to collaborative critique of poetry; similar ideas would apply for prose and, with further adaptation, other arts.\\

] \node (discovery) {\textbf{\emph{Discovery:}}}; % poet generates poem \node[single, right=8mm of discovery.east,text width=1.5cm] (poet) {\emph{poetry {\emph{text\\ generator}}; \node[single, right=4mm of poet.east] (poem) {P}; {T}; \draw [-latex] (poet.east) -- (poem.west); % critic listens to poem and offers feedback \node[single, right=4mm of poem.east,text width=1.5cm] (critic) {comment generator};

\subsection{Thought experiment: Serendipity by design} \label{sec:ww} To further evaluate our computational framework in usage, in this section we apply a thought experiment in system design, based around on a novel computational scenario where there is high potential for serendipity. As discussed above, sociological factors can influence serendipitous discoveries on a social scale. In our two case studies, user input played a significant role. The exploitation of social creativity and feedback can create scenarios where serendipity could occur for computers within a computer system as well.% In \cite{poetry-workshop}, we considered multi-agent systems that learn by sharing work in progress, and discussing partial understandings. The thought experiment we apply here explores serendipity in such scenarios, and is influenced by the ideas of \citeA{gabriel2002writer} on Writers Workshops. In \cite{poetry-workshop}, we described the preliminary designs for multi-agent systems that learn by sharing work in progress, and discussing partial understandings. % Following \citeA{gabriel2002writer} % we described a template for a pattern % language for interactions in a computational poetry workshop, closely

settings this is augmented with {\tt suggestions}. After any {\tt questions} from the author, the commentators may make {\tt replies} to offer clarification. %\footnote{We return to discuss further work with Writers Workshops and serendipity in Section \ref{sec:futurework}.} This is how these The key steps map quite conveniently into the diagram schematic description of serendipity that we introduced in Section \ref{sec:our-model}: \input{ww-schematic-tikz}

\begin{tikzpicture}[auto, node distance=2cm,>=latex'] \node [sum] (sum1) {}; \node [input, name=pinput, above left=.7cm left=.9cm and .7cm .9cm of sum1] (pinput) {}; \node [input, name=tinput, left=2cm left=2.2cm of sum1] (tinput) {}; \node [input, name=minput, below left of=sum1] (minput) {}; \node [input, name=minput, right of=sum1] (moutput) {}; \draw [->] (tinput) -- node{\vphantom{{\tiny g}}{\tiny \emph{presentation}}} node{\vphantom{{\footnotesize g}}{\footnotesize \emph{presentation}~~}} (sum1); \draw [->] (pinput) -- node{{\tiny node{{\footnotesize listening}} (sum1); \draw [->] (sum1) -- node{\vphantom{{\tiny g}}{\tiny node{\vphantom{{\footnotesize g}}{\footnotesize feedback}} (moutput); \end{tikzpicture} \hspace{1cm} \begin{tikzpicture}[auto, node distance=2cm,>=latex'] \node [sum] (sum1) {}; \node [input, name=pinput, above left=.7cm left=.9cm and .7cm .9cm of sum1] (pinput) {}; \node [input, name=tinput, left of=sum1] (tinput) {}; \node [input, name=minput, below left of=sum1] (minput) {}; \node [sum, right=1.5cm of sum1] (sum2) {}; \node [input, name=minput, right of=sum2] (moutput) {}; \draw [->] (tinput) -- node{\vphantom{{\tiny g}}{\tiny feedback}} node{\vphantom{{\footnotesize g}}{\footnotesize feedback~~}} (sum1); \draw [->] (pinput) -- node{{\tiny node{{\footnotesize \emph{questions}}} (sum1); \draw [->] (sum1) -- node{\vphantom{{\tiny g}}{\tiny node{\vphantom{{\footnotesize g}}{\footnotesize answers}} (sum2); \draw [->] (sum2) -- node{{\tiny node{{\footnotesize \emph{reflections}}} (moutput); \end{tikzpicture} \endgroup \end{center}