Content uploaded by Ana-Maria Olteteanu
Author content
All content in this area was uploaded by Ana-Maria Olteteanu on Aug 29, 2016
Content may be subject to copyright.
Towards an approach for the
computationally assisted creation of insight
problems in the practical object domain
Ana-Maria Oltet¸eanu1?
Abstract. Insight problems are creative solving problems used to assess
creativity in human participants, and empirically study insight-related
processes. However, not many such problems exist, and once a partici-
pant has been exposed to a particular problem, insight cannot be studied
anymore in that context. This paper proposes an approach for creating
more insight problems in the practical object uses domain. This approach
uses cognitive understanding of some of the processes that take part in
the solving of insight problems in this domain to create more such prob-
lems, and is amenable to computational implementation. The manual
creation of a couple of such problems with this approach is described.
Ways of making the approach computational are briefly discussed.
1 Introduction
Insight is an impressive phenomenon. Classical insight stories involve great artis-
tic or scientific achievements – like the ones presenting (i) Archimedes shouting
Eureka after having observed properties of water displacement that could help
him solve the crown problem, (ii) Watson dreaming of spiral staircases before
proposing the double helix structure of the DNA molecule and (iii) Kekul´e’s
daydream of an Ouroboros snake eating its own tail, before coming up with the
structure of the benzene molecule.
Insight, however, does not have to lead to big discoveries and involve historic
level creativity [Boden, 2003]. Insight presupposes seeing an existing problem
in a new way – a way in which it becomes solvable for the person seeing it.
Insight is studied empirically using various types of insight problems created
by humans. For example, [Dow and Mayer, 2004] have gathered a collection of
problems which they split into the following categories: mathematical, spatial
and verbal insight problems. A mathematical insight problem they mention is:
Which would be worth more, a pound of 10 dollar pure gold coins or half a
pound of 20 dollar pure gold coins; or would they be worth the same? Explain
your answer. A spatial insight problem gives the participant Figure 1 and gives
them the following task: without lifting your pencil from the paper, show how
you could join all 4 dots with 2 straight lines. An example of a verbal insight
problem from the same collection is the following: The legendary runner Flash
Fleetfoot was so fast that his friends said he could turn off the light switch and
?Corresponding author
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
2
jump into bed before the room got dark. On one occasion Flash proved he could
do it. How?
Fig. 1. The four dots problem
In this paper we will discuss a different (though at times overlapping) cat-
egory of insight problems – problems which involve the practical use of daily
objects, like the candle problem [Duncker, 1945] and the two strings problem
[Maier, 1931]1. This category generally involves objects and creative ways of
thinking about objects, their relations and their affordances, as well as cre-
ative modes of re-representing problems concerning objects and practical physi-
cal goals.
As fascinating as some of them might be, insight problems generally stop
being useful for the empirical study of the insight process if the participant
has encountered them even once beforehand. If the problem was once encoun-
tered, the participant might simply remember what the solution was, rather
than struggle to re-represent the problem, or experience a moment of sudden re-
representation. The experiment sessions involving problems which are already
known by the participants will thus be devoid of chances in gathering data
on authentic insight and creative problem solving processes. Considering that
one such problem can take a while to solve, and experimenters generally plan
for only a few such problems per experimental session, not having enough new
problems can result in lost experimental sessions, and the necessity of gathering
more empirical datapoints, to substitute for the ones in which a participant was
acquainted with one of the problems.
Within the practical object uses domain, not many insight problems exist,
and some of them are quite old problems, created by the minds alike those
of Maier and Duncker. A bigger dataset of such problems would benefit work
focused on the understanding of human creative cognition, and could further
find its application in creative problem solving work in robotics, and ambient
intelligence. Therefore, defining an approach to create more such problems, an
approach which could rely on computational means for problem creation assis-
tance, would be useful for both cognitive psychologists and for AI. We already
have some knowledge about the cognitive solving process of such problems. Cre-
ating them will benefit from such knowledge and bring about computational
applications for it.
This paper thus sets out to define an approach to creating insight problems
in the practical object uses domain, which is amenable to computational assis-
tance and/or implementation. This approach is cognitive in nature, looking at
1[Dow and Mayer, 2004] classify both of these as spatial problems. In the following
we will refer to such problems as insight problems in the practical object domain,
as to differentiate them from spatial problems involving abstract patterns, like the
one in Fig. 1.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
3
insight problems and their creation through the lens of insight related processes
explored in a previous cognitive theoretical frameowork of creative problem solv-
ing [Oltet¸eanu, 2016,Oltet¸eanu, 2014]. In the following, existing classical insight
problems will be analysed (Section 2); the approach will then be proposed (Sec-
tion 3); the cases of applying that approach to the creation of two such problems
will be explored (Section 4), and potential reliance on computational assistance
in problem creation with existing or future tools will be discussed (Section 5).
2 Classical insight problems – analysis
In this section, three examples of classical insight problems are taken as case
studies: (i) the candle problem, (ii) the two strings problem and (iii) the card-
board problem. These are analysed from the perspective of what makes them
insight problems, and what processes could have been used to create them. Prin-
ciples are then extracted for the creation of new problems. During this analysis,
we will assume the following:
(a) there is a simple, non-creative (or much less creative) version of the problem
which does not require insight;
(b) that by observing the steps and possible stumbling blocks of the solvers,
while they receive the version of the problem which does require creative skill,
we can extrapolate some processes which make a problem require creativity;
(c) that by using a variety of such interpolated processes and knowledge about
cognitive problem solving, we can begin to create more such problems, even
if we do not yet cover the entire variety of skills which can be tested through
them.
Candle problem
The candle problem by [Duncker, 1945] is stated as follows: You are given a
candle, a box of thumbtacks and a book of matches (see Fig. 2). You are supposed
to fix the lit candle unto the wall in a way that does not allow the wax to drip
below.
Fig. 2. The Candle problem
The candle problem is solved by taking out the thumbtacks from the box,
and using the box as a container and platform for the candle, then fixing it
(with thumbtacks) to the wall. The participants that get stuck while solving
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
4
this problem generally have trouble seeing the box of thumbtacks as a possible
container; this is sometimes helped if the box is empty.
Considering the types of processes involved in creative problem solving, in-
cluding re-representation, one can attempt to reverse engineer them, and check
whether this is conducive to the creation of similar problems. The following as-
pects can be synthesized from the candle problem to help the creation of new
problems:
– Hiding an object which is necessary for a (non-creative) solution of the problem
within a different other object. This aspect shows up in this problem by the fact
that a candle holder, which would have made the problem straightforward, is not
present. Such a candle holder needs to be re-represented out of existing problem
objects - from the perspective of the problem creator, one can look at this process
as one of hiding the candle holder in a different object (or set of objects) with
a similar affordance. Such a similar affordance however needs to be inferred cre-
atively, perhaps in a process similar to that used when solving the Alternative Uses
Test [Guilford et al., 1978].
– Hiding the affordance of an object by emphasizing a different affordance and (op-
tionally) having that affordance already taken up or in use. This aspect shows up
in the candle problem by the act of adding the thumbtacks inside the thumbtack
box2. Adding them near the box would have been a case for emphasizing the affor-
dance of the box as a container. Adding them within the box is a case for having
the affordance already in use. The purpose of the latter is, of course, to trigger and
thus help study the functional fixedness bias.
Two strings problem
The two strings problem by [Maier, 1931] presents to the participant a situ-
ation like the one in Fig. 3. The participant is told: A person is put in a room
that has two strings hanging from the ceiling. The task is to tie the two strings
together, but it is impossible to reach one string while holding the other.
The two strings problem is solved by making a pendulum from one of the two
strings and from a heavy object laid on the floor, like the pliers, then launching
one of the strings in pendular motion, as to be able to have it come on its own
towards one’s hand. The participants that get stuck when solving this problem
usually fail to see: (i) that the object can be set in motion on its own, rather
than requiring the motion of the solver; (ii) that the object could be created, as
this requires making it using other objects in the room.
Considering the steps involved in solving this problem, one can assume the
process of creating such a problem involves the following aspects:
– As a general strategy, make objects which need to be used for the solution lose
part of their affordances. A specific technique, observed in this case, is to enable
affordance loss by removing affordance related parts of the object. The object will
thus need reconstruction, while the parts themselves are less likely to trigger the
same affordance. In the two strings problem, this is done by removing the saliency
of the affordance of a pendulum to be mobile, through splitting the pendulum in
2It is the thumbtacks that make this a thumbtack box anyway, and one might be
biased to look at this box as a special container for the thumbtacks because of the
very description of the problem, which includes the verbal tag “thumbtack box”.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
5
Fig. 3. The Two Strings problem
two parts. A self-based motion frame of reference might be emphasized in the verbal
description of the problem (i.e. “it is impossible to reach a string while holding the
other”), which focuses the participant on thinking of themselves, rather than other
objects, as mobile.
– Split said object into parts and scatter the parts across the room. This is somewhat
overlapping in the context of this problem with the principle above. However, the
principle above can also involve removing solution-leading affordances of an object
in different ways, by, for example, changing the frame of reference, representing the
object in a space which makes the affordance less salient and constrains thinking
about the object.
– Hide the object required to solve the problem (or object parts) within other objects.
This principle is shared with the candle problem, with the added bonus that, here,
not just objects but also parts of objects are exchanged for similarly affording
objects. Here, the weight of the pendulum is hidden in heavy objects across the
floor.
– Add objects that lead to other possible affordances, and thus other possible con-
structions – in this case the chair (getting up on the chair gets triggered), the nails
(fixing one of the strings closer to the other by using the nails gets triggered).
Cardboard problem
The cardboard problem by [Duncker, 1945] is given as follows: You are asked
to help the experimenter attach this piece of cardboard to the loop on the ceiling.
How do you proceed? (our reconstruction of the depiction - Fig. 4).
This problem is solved by removing one of the paperclips, turning it around
to make an S-shaped hook, then using one end of the hook to pierce the corner
of the cardboard (the white paper corner can thus be kept in place too) and the
other end to attach to the loop. Most people attempt to solve this problem using
the nails and hammer on the table, various ways of standing the cardboard and
attaching it with nails to the loop.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
6
Fig. 4. The Cardboard problem
Creating problems like the cardboard problem could include the following
aspects:
– Hiding part of the solving objects as parts of other objects. In this case, the pa-
perclips are attached to the cardboard, and thus can be perceived as being part
of it, thus part of the object which needs attaching to the loop, rather than of the
tools which the attaching can be done with.
– Using one of the actions people would refrain from doing as part of the solution
– i.e. destroying an existing object, disobeying rules or norms or arrangements
perceived as implicit or unbreakable. In this case, not only the paperclips are
used to hold together the white piece of paper at the corner of the cardboard
(using them thus having the possible consequence of disassembling that part of
the object), but also piercing the cardboard can be viewed as a way of damaging
the cardboard irreversibly (in other problems, parts of objects can be pulled away
by disassembling the object to pieces, however the object can also be reassembled).
– Have other objects with a similar affordance in the scene, as to interfere with
finding the objects which would truly provide the solution. In the case of this
problem, the affordance that nails and hammer have to fix something to a wall or
a ceiling interferes with seeing the less salient affordance of the paperclip, setting
the nails and hammer center stage as red herrings.
3 Approach
Various processes of solving seem necessary for the above described problems,
and thus various principles of creating more such problems become apparent
after this analysis. In this section, the extracted principles are discussed and
summarized in an approach.
As mentioned in section 2, one of the parts of our analysis assumed that there
is a simple, non-creative version of the problem, which does not require insight.
For example, the candle problem would be in its simpler form if it would offer
a candle holder as part of the existing objects, rather than the thumbtack box;
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
7
the two strings problem would be in its simpler form if one of the strings would
be a pendulum already; the cardboard problem – if the paperclip was already
twisted in the necessary S shape, or at least detached from the cardboard.
Consequently, this approach starts from the assumption that more creative
problems can be constructed in the practical object domain, by taking simple
day to day non-creative problems, considering one of their normal solutions, and
then hiding the possibility of applying that solution from the solver via a set of
re-representations and creative uses of objects (which then have to be traversed
back by the solver) and problem templates (or sets of viable action plans).
A non-exhaustive list of some of the techniques that can thus be used in
problem creation, in light of the previous case studies, includes:
(i) Diminishing the saliency of the objects required for the solution, by (a)
putting them in a different context of affordances and possibly (b) having
those affordances already allocated or used;
(ii) Hiding objects in a different form – by re-representing them as other ob-
jects which have similar properties and affordances (but for which said
affordances might not be as salient as for the initial objects);
(iii) Decomposing the solution in different parts, and re-representing the parts
in different structures or objects;
(iv) Representing needed parts as integrated parts of other objects;
(v) Using an object twice in the solution, with two different contexts of affor-
dances. In this case, participants need to look at both sets of affordance
contexts, to perceive the object in both of its potential roles, similar to
being able to look at two ambiguous figures the perception of which can
emerge from the same set of elements;
(vi) Adding to the problem other salient objects, the affordances of which might
interfere with the solution;
(vii) Making use of natural or learned biases against breaking objects, crossing
commonsense or common practice norms or aesthetic values.
Part of the techniques in this approach could be loosely summarized as the
following (obj stands for object, af f for affordance, P T for problem template3,
sol for solution and simP rop for similar properties):
1. Embed in different affordance contexts:
If (objx∈P Tsol)∧(∃af f (objx)6=af fsol(objx)∨
∃aff(objx, objk)6=af fsol (objx)∨
∃P Tx|objx∈P Tx, af f(P Tx)∩af fsol ) = 0)
then display(af f (objx)|af f (objx)6=affsol(objx)∨
display(af f (objx, objk)) ∨
display(elementsOf(P Tx))
2. Use creative object replacement:
If objx∈sol ∧ ∃obja|simP rop(objx, obja)∨simP rop(objx, par tOf(obja))
Then replace(objx, obja)∨replace(objx, partOf (obja))
3A problem template is a set of actions that will lead to a particular solution or
affordance; such sets of actions are part of the commonsense knowledge of the subject.
Creative use of problem templates is detailed in other works [Oltet¸eanu, 2014].
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
8
3. Decompose object: decompose(objx) = partsOf (objx)
4. Represent needed parts as parts of other objects:
If objx∈P Tsol ∧partOf (objb, objx)∧objb/∈problem →replace(objx, objb)
5. Double use:
If objx∈P Tsol ∧obja∈P Tsol ∧simP rop(objx, obja)→remove(objx)
6. Adding other salient objects or templates:
If ∃affx|sim(affx, af fsol)∧affx/∈P Tsol ∧affx∈P Tx, P Tx/∈P Tsol
show(af fx)∨show(objk1, objk2...objkn ∈P Tx)
4 Creating new insight problems – two cases
In this section, two problems that were created using the above principles are
discussed. These are: (i) the blown away teddy problem and (ii) the Jack and
Jill weight problem.
The blown away teddy problem
The blown away teddy problem presents the participant with the following
task: The wind blew your son’s teddy bear from the clothesline into your neigh-
bour’s garden. The neighbour is in holidays and the fence is too high to climb.
How can you retrieve the teddy? Fig. 5 shows the problem.
Fig. 5. The blown away teddy problem
This problem is solved by constructing a fishing rod, using the mop, the
clothesline, and a clothes hanger attached to the clothesline.
The problem was constructed using the approach proposed in Section 3, in
the following steps:
1. Start from a problem and an existing solution. The problem is that you need
to obtain an object that is far away. The solution is to reach for the object.
2. Make solution creative. A fishing rod is used for fishing, but can have the
alternative use of hooking something that is far away. The object replacement
part of a system like OROC [Oltet¸eanu and Falomir, 2016] can be used to
generate alternative uses.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
9
3. Decompose object into parts, so that it will require composition. The parts
of the fishing rod used here are the pole, the string and the hook.
4. Hide the parts of the solution in other objects with different salient affor-
dances, or in parts of other objects. The string is presented as a clothesline
(one to one object mapping). The hook is presented as part of the clothes
hanger (part of object). The pole is presented as part of the mop (part of
object). The participant can also attempt to use the pole that is part of the
umbrella, if the participant perceives this as movable.
5. Embed the replacement (or re-represented) objects in different contexts of
affordances. The clothesline is presented holding clothes to dry, attached to
the umbrella pole. The hangers have clothes on them, and are also partly
obscured visually by the clothes (the part which provides the affordance
conducive to the solution is, however, visible). The mop is presented next to
the bucket and a water puddle, which emphasize its cleaning affordances.
Parts of the problem are also ambiguous – how tall the fence is, the height
of the table, the distance to the teddy. The experimenter can let the participant
attempt various strategies within this ambiguity, to observe various types of
constructions created, paths pursued and forms of creative reasoning. Other
constraints can then be set in place and communicated to the solver (e.g. the
teddy is too far to be reached just using the mop) in order to observe new
strategies at play.
This process can be summarized as the following:
1. Initial problem:
(a) Starting condition: faraway(subj ect, teddy).
(b) Goal: has(subject, teddy).
(c) Solution: reach(subject, teddy )
(d) Starting template: reach(subject, teddy )→has(subject, teddy)
2. Creative version of a reach template:
reach(subject, teddy)→f ish(f ishing rod, teddy); beach, r iver, pool /∈problem4
3. Decompose:
decompose(fishing rod) = (pole, string, hook)
4. Creative replacement:
(a) creative replacement(string) = clothesline
(b) creative replacement(pole) = mop handle
(c) creative replacement(hook) = clothes hanger
5. Embed in affordances or used affordances:
(a) embed(clothesline)→show(attached(clothesl ine, pole), on(clothesline, clothes))
(b) embed(mop handle)→embed(mop)→show(nextT o(mop, bucket))
show(mopping(mop, w ater))
(c) embed(clothes hanger)→show(on(clothes hanger, clothes)),
show(on(clothesline, clothes hang er))
4This means we have to avoid a fishing related context for the rest of the scene, so
the scene cannot be of a subset of scenes that might trigger fishing templates, like
beach and river, nor scenes that are similar to them – like pool.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
10
Jack and Jill weight problem
The Jack and Jill weight problem presents to the participant a situation like
the one displayed in Fig. 6. The participant is given the following task: Jack and
Jill are arguing about whom weighs more. What could they do to find out for
certain?
Fig. 6. The Jack and Jill weight problem
This problem is solved by making a seesaw from the bucket and a surfing
board and placing Jack and Jill at opposite ends.
The problem was constructed using the following steps and techniques from
the approach:
1. Start from a problem and an existing solution. The initial problem template
was that balancing scales are used to measure weight.
2. Change the problem so that the solution would be creative – use seesaw
instead of scales; put the problem in the beach setting (change of context
to one which less affords thinking about weights and balancing scales, like a
kitchen), and use of a different object with similar properties.
3. Split the object in various parts – the seesaw was split into a pivot and a
support plank.
4. Hide the objects which form the solution by re-representing them as other
objects or object parts – the plank was turned into a surfboard (also adapta-
tion to the current context) and the pivot into a bucket (similar adaptation).
Through the adaptation to context, both objects thus can be envisaged as
belonging to a normal beach scene, rather than triggering the attention of the
participant as objects that have especially been added to the beach context
because they are part of the solution.
5. Hide the objects in different contexts of affordances and possibly have those
affordances be already in use – the bucket is turned with its side up, and in
its container affordance (full of sand); the surfboard is being surfed on, and
quite far away, which makes it visually less salient. Some participants might
also have social qualms with solution steps that involve asking for an object
which is being currently used and belongs to someone else.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
11
6. Add objects which act as red herrings, providing a similar affordance as the
one necessary for solving the problem, thus getting the participant started
on a different track – the small plastic swimming pool can get the participant
started in this case on an Archimede’s principle template.
This process can be summarized as the following:
1. Initial problem defined:
(a) Starting condition: unknownW eight(x, y ).
(b) Goal: f indW eightDif fer ence(x, y).
(c) Solution: balance(x, y)
(d) Starting template: balance(x, y)→f indW eightDif f erence(x, y)
2. Creatively change the initial template (problem+solution):
balance(x, y)→seesaw(personx, persony), P Tx6=kitchen
3. Decompose:
decompose(seesaw) = (pivot, suppor t pl ank)
4. Creative replacement:
(a) creative replacement(pivot) = bucket
(b) creative replacement(plank) = surfboard
5. Embed in affordances or used affordances:
(a) embed(bucket)→show(in(bucket, sand), near(bucket, toy spade))5
(b) embed(surf board)→show(on(water, surf board)), on(surf board, surf er))
6. Addition of red herring objects:
sim(measure weight, measure volume)∧measure volume /∈PTsol ∧
measure volume ∈P TArchimedes ∧bathtub ∈P TArchimedes ∧
sim(bathtub, pool with water )→show(pool with water)
An extra point can be made about embedding the entire problem in a new
contextual setting. The problem template of balance is in this case creatively
transformed to a template about seesaws, and then imported in the contextual
setting of a beach. The beach location could have been chosen as a consequence
of having already picked one of the two replacement objects for the seesaw parts –
the surfboard. Laying this object in its own contextual setting made the natural
choice a beach, and the replacement for the second object, which did not have
that many constraining properties, could be fixed to another object in the same
setting – the bucket.
5 Discussion – towards a computational approach
As shown in the above test cases: (i) insight problems in the practical uses of ob-
jects domain can be analysed from the cognitive perspective of re-representation,
and creative inference, (ii) such principles can be put together in an approach
towards creating more insight problems and (iii) the approach can be used to
create more insight problems in the specified domain.
5The PT of using a bucket to dig and play with sand is supported by the neighborhood
of objects such a spade, a sand castle, etc.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
12
This approach might not reflect a reverse engineering of all the types of
insight processes, or might not yet result in all the types of problems that can be
created in the domain. However, starting from creating some such problems and
evaluating them with human participants will have the impact of understanding
and controlling the process of creating insight problems more thoroughly, and
thus, in the future, providing a wider database of problems, based perhaps on a
wider array of processes.
Creating insight problems might seem like a lofty computational pursuit.
However, as shown above, the processes implied by this approach are substantial
enough to allow for formalization. An interesting next step would be to tackle
the issue using computational assistance when creating such problems. Part of
the tools needed for such an approach already exist, at least in prototype form.
Take item (ii) of the list of techniques provided by the approach here – hiding
objects in a different form – by replacing them with objects which have similar
properties and affordances (but for which solution related affordances might not
be as salient as for the initial objects). The creative object replacement (OR)
part of OROC [Oltet¸eanu and Falomir, 2016] can be used to generate items from
the practical objects domain which have similar affordances as the initial items.
OROC makes creative inferences about new affordances of known objects, based
on the similarity between said objects to other objects on shape and material
properties. The object composition (OC) part of the same system can in part
take care of items (iii) and (iv) on the list, specifically by decomposing various
objects which are part of the solution in object parts, then finding similar objects
with similar properties (OR) in its knowledge base; these objects, or the ones
they are part of, can be used to substitute initial objects which are a salient part
of the solution.
Other parts of this approach, like (i), (v) and creative transfer of a simple
problem can be based on OROC knowledge, but also require knowledge of prob-
lem templates. Such knowledge should include contexts of affordances in which
various objects get engaged, functional subsets of objects which are employed
in such templates, qualitative or quantitative measures of similarity of template
affordance, and some measure for when templates achieve similar but not quite
the same results – thus interfering with the human judgement because of the
similarity component, but not being able to help participants solve the problem.
In conclusion, a few case studies of classical insight problems have been anal-
ysed, in order to extract a set of principles which can be used in the creation
of new insight problems. A non-exhaustive approach towards mechanisms that
can be used to create such problems, and that is amenable to computational
implementation has been proposed. Then, the applicability of the approach has
been analysed, by describing it in the context of two newly created insight prob-
lems. Some existing tools that can be used to assist with this process have been
briefly discussed. As future work, we plan to start (i) implementing this ap-
proach and/or relying on computational assistance, and (ii) experiment with
problem template acquisition, problem template transfer and creative replace-
ment of problem templates.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
13
References
[Boden, 2003] Boden, M. (2003). The Creative Mind: Myths and Mechanisms. Rout-
ledge.
[Dow and Mayer, 2004] Dow, G. T. and Mayer, R. E. (2004). Teaching students to
solve insight problems: Evidence for domain specificity in creativity training. Cre-
ativity Research Journal, 16(4):389–398.
[Duncker, 1945] Duncker, K. (1945). On problem solving. Psychological Monographs,
58(5, Whole No.270).
[Guilford et al., 1978] Guilford, J., Christensen, P., Merrifield, P., and Wilson, R.
(1978). Alternate uses: Manual of instructions and interpretation. Orange, CA:
Sheridan Psychological Services.
[Maier, 1931] Maier, N. R. (1931). Reasoning in humans. II. The solution of a problem
and its appearance in consciousness. Journal of comparative Psychology, 12(2):181.
[Oltet¸eanu, 2014] Oltet¸eanu, A.-M. (2014). Two general classes in creative problem-
solving? An account based on the cognitive processes involved in the problem struc-
ture - representation structure relationship. In Besold, T., K¨uhnberger, K.-U., Schor-
lemmer, M., and Smaill, A., editors, Proceedings of the Workshop on Computational
Creativity, Concept Invention, and General Intelligence, volume 01-2014 of Publica-
tions of the Institute of Cognitive Science. Institute of Cognitive Science Osnabr¨uck.
[Oltet¸eanu and Falomir, 2016] Oltet¸eanu, A.-M. and Falomir, Z. (2016). Object re-
placement and object composition in a creative cognitive system. Towards a compu-
tational solver of the Alternative Uses Test. Cognitive Systems Research, 39:15–32.
[Oltet¸eanu, 2016] Oltet¸eanu, A.-M. (2016). From simple machines to Eureka in four
not-so-easy steps. Towards creative visuospatial intelligence. In M¨uller, V., editor,
Fundamental Issues of Artificial Intelligence, volume 376 of Synthese Library, pages
159–180. Springer.
Oltețeanu, Ana-Maria - Towards an approach for computationally assisted creation of insight problems in the practical object domain,
in Proceedings of the 5th International Workshop on “Computational Creativity, Concept Invention, and General Intelligence“,
editors Besold, T.;Kutz, O.; and Leon, C., Publications of the Institute of Cognitive Science, Osnabrck, 2016.
