\label{plan-of-action}
(Joe, Alison): Significantly clarify the argument and summarise it in the introduction.
We are offering one possible computational definition of serendipity
Serendipity is not the same as luck. It’s a matter of learning something, in a way that’s unanticipated. Looking for something and finding something else.
Explain the aspects of the model better, e.g. why is it essential that the trigger is not under the control of the system.
Clearly summarise the offering of the paper.
(Joe): Move our formal definition of serendipity (e.g. the diagram) up to meet the literature review, as a new section ‘Formal definition of Serendipity’. (It’s a key contribution of the paper.)
We will clearly connect the heuristic criteria from Alison with the figure.
In addition a quick graphical summary of the 13 criteria
(Joe): Drop sections 3 and 4, and move key concepts to “future work”
Section 3 (FloWr) - heavily condense and put into future work (some overview of Joe’s concrete implementation plans). Explain with minimal references.
Section 4 (Design patterns) - heavily condensed - “Just So Stories” paragraph in Section 5.3 as a potential application. Explain some history about design patterns and say that, for serendipity, the question is where do new “design” ideas come from. (I.e. discovery of a new approach.) But make this future work.
“We are highlighting how design patterns and the other ideas in this paper could be used to build a context where serendipity will take place.”
(Anna): Remove Section 5.3 (save it for another paper about Writers Workshops). It’s relevant for “embedded creativity” but “Writers Workshops” themselves can be a footnote. The actual idea here is more general.
Anna can add more about evaluation in the creative process
The idea of the WW (or just social revision) is an example of a place where serendipity can occur.
(Anna): Leading into our thought experiment: “An emerging theme in computing is exploitation of social creativity and feedback. Our computational model contributes to theorising this work.”
Include another example with computational serendipity? Maybe the example from Kaz’s thesis
It would not be hard to find an example of a music system noticing that a note was wrong and playing. Make sure we include at least one example that is not “technically improbable” – better to include several that have been realised (e.g. Copycat)
(Christian): Copycat or any other historical examples of serendipity in computing, or explanation of why there are none (argue for or against, in the background section, as a new §§, and perhaps again later in the document as a further analysis to accompany our thought experiment).
Concrete lower bound examples and counterexamples, e.g. would it be possible for “merely generative” systems to exhibit serendipity? – case of genetic algorithms
What is the difference between serendipity and good luck? (E.g. a random act that leads to an outcome that is evaluated positively.)
What are the strict requirements and what are only the supportive factors that make serendipity “likely”? Or is it a matter of degree?
Is it the case that serendipitous systems would be more ‘sagacious’ in recognizing interesting triggers? - explain, especially in connection with computational search.
What about ‘regular’ systems that work by applying inference procedures on symbolic representations to yield new representations?
e.g. theorem provers
Evaluate existing approaches to “computational learning” - are they serendipitous?
(Simon, Alison): Clarify the extent to which serendipity is something that “actually exists” or is something that is only perceived to exist.
It does not seem to be an “essentially contested concept”, just a potentially confusing one. One contribution of the paper is to clarify this.
Clarify the relationship to other key concepts in computational creativity / creative computing
(Alison): Include a section early on that defines any other keywords that we refer to later, like the word “dynamic”.
(Alison): Improve exposition of the analysis of Pek van Andel’s patterns.
What do we hope to achieve with this analysis, and our diagram?
Have we done the analysis in some verifiable way, i.e. “where does the analysis come from (i.e. which aspect occurs in which pattern)? Is there clear consensus on this?”
(Joe, all): Make referencing less intensive for the reader.
Use APA style referencing and cut down on number of references.
Clearly explain in narrative form what sort of literature we will draw on.
Perhaps the historical examples of serendipity should be confined to a separate “recommended reading” section and not referenced directly in the text.
(Christian, Anna): Shorten and improve the literature review.
Preserve key features of the general survey, but include a more thorough review of recent related work in computing, including work in the Cognitive Computation journal.
There has been prior work on surprise (Mary Lou Maher + Kazjon Grace - https://www.aaai.org/ocs/index.php/WS/AAAIW14/paper/view/8779 and also their paper at ICCC 2013 or 2014) and discovery (Kaz’s AAAI paper)
(Joe): Confine philosophy references (e.g. Bergson, Deleuze) to the background section so that it doesn’t confuse anyone about what we’re actually offering in the paper.
Don’t refer to them in the conclusion, but do summarise the contribution of this paper again in the conclusion (hint: it should be what we say in the title).
Re-summarise again in the abstract.