Authorea

Joe Corneli revise wording around SPECS, add example, and continue with CC example about 9 years ago

Commit id: f6bb3205abbf7f9147f6ef635b9a276756519402

deletions | additions

consisting of previous experiences, background knowledge, a store of unsolved problems, skills, expectations, and (optionally) a current focus or goal. It then processes a \emph{\textbf{serendipity trigger}} that is at least partially the result of factors outside of its control, including randomness or unexpected events. The system then uses reasoning techniques and/or social or otherwise externally enacted alternatives to create a \emph{\textbf{bridge}} from the trigger to a result. The \emph{\textbf{result}} is evaluated as useful, by the system and/or by an external source.} \item[\emph{(\textbf{B - Dimensions})}] \emph{Serendipity, and its various dimensions, can be present to a greater or lesser degree. If the criteria above have been met, we generate ratings as estimated probabilitiesin $[0,1]$, along several dimensions: % \emph{($\mathbf{a}$ - \textbf{chance})} how likely was this trigger to appear to the system?

results.}\end{quote} We will develop several examples of the application of this framework to examples from computing in Section \ref{sec:computational-serendipity}. First, a quick illustration using one of the historical examples mentioned earlier. \begin{mdframed} \vspace{-.35cm} \paragraph{De Mestral's invention of Velco\texttrademark}~\\[.05cm] \begin{description} \item[\emph{\textbf{prepared mind}}] De Mestral was a trained engineer and experienced inventor. \item[\emph{\textbf{serendipity trigger}}] Burdock burs stuck to clothes and fur. \item[\emph{\textbf{bridge}}] Viewing the burs under a microscope, followed by experiments to create a fastener that worked on the same principle. \item[\emph{\textbf{result}}] Velcro\texttrademark \item[\emph{\textbf{chance}}] High chance of encountering burs while out walking. \item[\emph{\textbf{curiosity}}] Few people would investigate them under a microscope. \item[\emph{\textbf{sagacity}}] Few people would conceive of a useful application or have the perseverance required to carry out product development. \item[\emph{\textbf{value}}] High; hook-and-loop fasteners are widely used. \begin{itemize} \item Ruling: ``highly serendipitous''. \end{itemize} \item[\emph{\textbf{dynamic world}}] New fabrication principles required. \item[\emph{\textbf{multiple contexts}}] Biomimicry (art imitating life). \item[\emph{\textbf{multiple tasks}}] N/A. \item[\emph{\textbf{multiple influences}}] N/A. \end{description} \end{mdframed} %% In order to develop connections with our theoretical framework, and %% because existing experiments have not been particularly strong, we

@book{colton2002automated, title={Automated theory formation in pure mathematics}, author={Colton, Simon}, year={2002}, publisher={Springer} } @inproceedings{abbassi2009getting, title={Getting recommender systems to think outside the box}, author={Abbassi, Zeinab and Amer-Yahia, Sihem and Lakshmanan, Laks VS and Vassilvitskii, Sergei and Yu, Cong},

\section{Serendipity ina computational context} systems} \label{sec:computational-serendipity} The 13 criteria from Section \ref{sec:literature-review} specify the conditions and preconditions that are conducive to serendipitous discovery. Here, we revisit each These criteria have been further formalised in Section \ref{specs-overview}. % \citeA{pease2013discussion} used a slightly different version of these the SPECS criteria to discuss three examples of serendipitous behaviour: in dynamic investigation problems, model generation, and briefly summarise how they can be thought about from a computational point poetry flowcharts. Two additional examples using the revised criteria are described below. These example serve the purpose of illustrating our revised criteria, and also show forays of view, again focusing on examples. computational intelligence into domains known for serendipity in their everyday cultural context. We then present turn to a more elaborated thought experiment that evaluates the these ideasdescribed above in the course of developing a new system design. % \input{writers-workshop-background-long} \subsection{Prior partial examples} \citeA{pease2013discussion} used a somewhat different version of the SPECS criteria to discuss three examples, related to dynamic investigation problems, model generation, and poetry flowcharts. The \paragraph{{[}To add: Jazz.{]}} % \paragraph{{[}To add: HR.{]}}

of the primary contexts in computing where serendipity is seen to play a role. As we noted, these systems mostly focus on discovery. Nevertheless, certain architectures that also take account of invention may would match the all of criteria described by our model. We Here we draw on the observation that recommender systems notonly aim to \emph{stimulate} serendipitous discovery for the user: they also have the task of \emph{simulating} when this is likely to occur. A recommendation is typically provided if the system suspects that the item will be likely to introduce ideas that are close to what the user

new approach that helps to address these goals better. From the perspective of our model, \textbf{chance} will only have a significant role if when the system has the capacity to learn from user behaviour. Indeed, In fact, Bayesian methods are used in contemporary recommender systems (surveyed in Chapter 3 of \citeNP{shengbo-guo-thesis}). The typical commercial perspective on recommendations is related to the process of ``conversion'' -- turning recommendations into clicks and clicks into purchases. 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. The typical commercial perspective on recommendations is related to users, potentially at the process cost of ``conversion'' -- turning recommendations into clicks and clicks into purchases. recommendation quality. Measures of \textbf{sagacity} would relate to the system's ability to draw inferences from user behaviour to that would update the recommendation model. For example, the system might do A/B testing to decide how novel recommendation strategies influences influence conversion. The \textbf{value} ofnew recommendation strategies can be measured in terms of traditional business metrics or other organisational objectives. A \textbf{dynamic world} is a precondition for 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 based on the results of these experiments. 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.

related to poetry (e.g.~definitions of words, valence of sentiments, metre, repetition, density, etc.) and code (e.g.~the data, functions, and macros in which the poetic concepts and workshop protocols are embodied). Some notable previous early experiments with concept invention have been fraught with questions about autonomy \cite{ritchie1984case,lenat1984and}. \textbf{[Some comment about HR here?]} \citeA{colton2002automated} presented a system that was convincingly autonomous -- the system was able to generate novel conjectures that surprised its author. However, \citeA{pease2013discussion} note that this system was not convincingly serendipitous: ``we had to willingly make the system less effective to encourage incidents which onto which we might project the word serendipity.'' One cognitively inspired hypothesisis that could describe the serendipitous formation of new concepts is the notion that the development of new concepts is closely related to formation 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 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 on the poem -- simply describing what's 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 of the

\subsection{Thought experiment: Serendipity by design} \label{sec:ww} To evaluate our computational framework in usage, we apply a thought experiment based around a novel computational scenario where there is high potential for serendipity. As discussed above, sociological factors can influence serendipitous discoveries on a social scale. The exploitation of social creativity and feedback can create scenarios where serendipity could occur. occur for computers 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