Authorea

Joe Corneli massaging text to match memo about 9 years ago

Commit id: f1ed40227556a302ad9673ffac7d4f0030631921

deletions | additions

%% Saved with string encoding Unicode (UTF-8) @article{wu2013curiosity, title={{C}uriosity: {F}rom psychology to computation}, author={Wu, Qiong and Miao, Chunyan}, journal={ACM Computing Surveys (CSUR)}, volume={46}, number={2}, pages={18}, year={2013}, publisher={ACM} } @inproceedings{grace2014using, title={{U}sing {C}omputational {C}reativity to {G}uide {D}ata-{I}ntensive {S}cientific {D}iscovery}, author={Grace, Kazjon and Maher, Mary Lou}, booktitle={{W}orkshops at the {T}wenty-{E}ighth {AAAI} {C}onference on {A}rtificial {I}ntelligence}, editor={Yolanda Gil and Haym Hirsh}, year={2014}, note={Discovery Informatics Workshop: Science Challenges for Intelligent Systems.} } @article{bishop-erden-special-issue, title={{C}omputational {C}reativity, {I}ntelligence and {A}utonomy {[}{S}pecial issue{]}}, editor={Bishop, Mark J. and Erden, Yasemin J.},

Combined with the ability to learn, \textbf{curiosity} could be described as the urge to make ``outside-the-box''\footnote{\citeA{abbassi2009getting}.} recommendations specifically for the purposes of learning 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 inferences from user behaviour. For example, the system might do A/B testing to decide how novel recommendation strategies influence conversion. Again, currently this would typically be organised by the system's developers. The \textbf{value} of recommendation strategies can be measured in terms of traditional business metrics or other organisational objectives. A \textbf{dynamic world} which nevertheless exhibits some regularity is 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. % As a general comment, we would say that this is largely how % \emph{research and development} of recommender systems works, but % without the same levels of system automony envisioned here. %%% \begin{table}[ht!] {\centering \renewcommand{\arraystretch}{1.5} \footnotesize

\caption{Summary: applying computational serendipity model to positive case studies\label{caseStudies}} \end{table}% \normalsize %%% recommendations specifically for the purposes of learning 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 inferences from user behaviour. For example, the system might do A/B testing to decide how novel recommendation strategies influence conversion. Again, currently this would typically be organised by the system's developers. The \textbf{value} of recommendation strategies can be measured in terms of traditional business metrics or other organisational objectives. A \textbf{dynamic world} which nevertheless exhibits some regularity is 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. % As a general comment, we would say that this is largely how % \emph{research and development} of recommender systems works, but % without the same levels of system automony envisioned here.

never have happened.}''~\cite{bergson2010creative} \end{quote} As we have indicated, serendipity would seem to require features of both; both discovery and invention; 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.

\subsection{Future Work} \label{sec:futurework} \label{sec:hatching} In looking for ways to manage and encourage serendipity, we are drawn to the approach taken by the \emph{design pattern} community \cite{alexander1999origins}. The essential features of this approach are described below, but weshould point out straight away that we propose to use design patterns in rather nonstandard fashion: fashion. These adaptations to the typical design pattern methodology are proposed to parallel the four themes outlined above. \begin{itemize} \item[(1)] We want to encode our design patterns directly in runnable programs, not just give them to programmers as heuristic guidance. \item[(2)] We want the (automated) programmer to generate new design patterns, not just apply or adapt old ones. \item[(3)] We want our design patterns patterns, working in combination, to help find new problems, emergent problems and patterns, not just capture the solutions to existing ones. known problems. \item[(4)] We want our design patterns to play an overt role in the dynamical systems they describe. \end{itemize} \citeA{meszaros1998pattern} describe the typical scenario for authors of design

models personal motivations, social interactions and the evolution of domains.'' Paper 10, d'Inverno and Luck's \citeyear{d2012creativity} ``Creativity Through Autonomy and Interaction'', also contains a theoretical engagement with these themes, and presents a formalism for multi-agent systems that could usefully be adapted to model serendipitous encounters. Both papers are particularly concerned with \emph{motivation}, a theme topic that relates to our notion of a both the prepared mind andto the topic theme of embedded evaluation. We believe that our clarifications to the multifaceted concept of serendipity will help encourage future computer-aided (and

An active research community investigating computational models of serendipity exists in the field of information retrieval, and specifically, in recommender systems \cite{Toms2000}. In this domain, \citeA{Herlocker2004} and \citeA{McNee2006} view serendipity as an important factor for user satisfaction, alongside accuracy and diversity. Serendipity in recommendations variously require the system to deliver an \emph{unexpected} and \emph{useful}, \emph{interesting}, \emph{attractive} or \emph{relevant} item. % \cite{Herlocker2004} \cite{Lu2012},\cite{Ge2010}. Definitions differ as to the requirement of \emph{novelty}; \citeA{Adamopoulos2011}, for example, describe systems that suggest items that may already be known, but are still unexpected in the current context. While standardized measures such as the $F_1$-score or the (R)MSE are used to determine the \emph{accuracy} of a recommendation (as very close to what the user is known to prefer), there is no common agreement on a measure for serendipity yet, although there are several proposals \cite{Murakami2008, Adamopoulos2011, McCay-Peet2011,iaquinta2010can}. In terms of our model, these systems focus mainly on producing a \emph{serendipity trigger} for the user, user and in support of discovery, but they include aspects of user modeling which could bring other elements into play, as we will discuss in Section \ref{sec:computational-serendipity}. Recent work has examined related topics of \emph{curiosity} \cite{wu2013curiosity} and \emph{surprise} \cite{grace2014using} in computing. This latter work seeks to ``adopt methods from the field of computational creativity [$\ldots$] to the generation of scientific hypotheses.'' In contrast to the typical application of recommender systems, this is an example of an effort focused on computational invention. Paul Andr{\'e} et al.~\citeyear{andre2009discovery} have examined serendipity from a design perspective. These Like us, these authors also propose proposed a two-part model, in which what we have called \emph{discovery} above exposes model encompassing ``the chance encountering of information, and the unexpected, while \emph{invention} is sagacity to derive insight from the responsibility another subsystem that finds applications. encounter.'' According to Andr\'e et al., the first phase is the one that has most frequently been automated -- but they suggest that computational systems should be developed that support both aspects. Their specific suggestions They specifically suggest to focus on representational features: \emph{domain expertise} and a \emph{common language model}. Although tremendously useful when they are available, these features are not always enough to account for serendipitious events. Using the terminology we introduced earlier, these features seem to exemplify aspects of the \emph{prepared mind}. However, as we mentioned above, the \emph{bridge} is a distinct process that mental preparation can support, but that it does not always necessarily fully determine. For example, participants in a poetry workshop may possess a very limited understanding of each other's aims or of the work they are critiquing, and may as a consequence talk past one another to a greater or lesser degree --

\emph{asymmetric partial knowledge} can support serendipitious findings. These examples suggest that a distinction between emergent knowledge of other actors and knowledge about an underlying domain may be useful -- although the distinction would somewhat may be less relevant if the underlying domain itself has dynamic and emergent features. \emph{Social coordination} among human users of information systems is a current research topic. \citeA{rubin2010everyday} point out that

\bigskip \noindent 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

\emph{reflections}) are the responsibilities of the presenting author; the other elements (listening, feedback, and answers) are the responsibilities of the attendant critics. % The system as a whole can be further decomposed into generative components as follows: \bigskip \begin{figure} {\centering \input{ww-generative-tikz} \par} \caption{Generative schematic for a Writers Workshop\label{fig:generative-diagram}} \end{figure} % The system as a whole can be further decomposed into generative components, as in Figure \ref{fig:generative-diagram}.