Problem Solving Flashcards
what is problem solving
there is a clearly defined goal but the path to this goal is uncertain
it takes conscious planning to arrive this goal
what are problem spaces (Newell & Simon, 1972)
conceptualised problems as a search process through a space
the problem is defined as an initial state
goal is to move through various intermediate states via operators to move from one state to another
what is the tower of Hanoi
goal is to move all three disks to the third peg but can only move one disk at a time and cannot place a larger disk on top of a smaller disk
to solve this people must construct a search space
what is the algorithmic approach to searching a problem space
a procedure guaranteed to solve a problem
requires the individual to plan though every possibility
however the process is inefficient, slow and intractable
what is bounded rationality
People think and behave rationally
within the limits imposed by their
cognitive capacities, available time,
and access to information.
we use simple heuristics to get around these limits
these allow us to solve a range of problems simply
what are multi-step problems
problems with clear structures but complex solutions e.g tower of hanoi
we can measure solution times and errors or analyse verbal protocols
what are the various heuristics we might use in computer models
simplify process of search so that we are not representing every possible state
depth-first search: pick a path and attempt to move through the space and see if this leads to a possible solution
breadth-first search: what are all the possible moves at the first step, then given the first step what are all the possible moves at the second step
what are the various heuristics we might use in human models
hill climbing: initial state –> pick a move that makes current state more/most similar to goal, i.e. take a step that moves you upwards, should reach goal.
however can reach local maxima (Thomas, 1974)
people struggle when there are moves that go against hill climbing heuristic
Despite getting stuck at these points, people are able to solve problems that involve violations of hill climbing suggesting that this is not the only algorithm used
how do people overcome local maximum problems
by breaking goals down into smaller steps
what is means-ends analysis
like hillclimbing in that you genenrally try to make moves that get you close to the goal state, however there is the addition of subgoals for local maximum problems that allow you to do something when you get stuck
at every state identify difference between current state and goal state
is there is set a subgoal to eliminate this difference
continues iteratively until there is no difference
the subgoal consists of an iterative algorithm: is there an operator to reduce difference - apply, if you can, if not apply a new subgoal
what are the two key processes in problem solving
Choosing operators/ actions that get you closer to the goal
If you get stuck, you set the goal of getting unstuck and being able to use the operator you’d like to
In an interactive process
what evidence is there for means-ends analysis
working solution to many problems - the general problem solver able to produce solutions to many problems when implemented in AI (not necessarily the way people think)
matches verbal protocols as people are solving problem - show division into goals and subgoals and verbal protocol is a solution to the problem when reconstructed
these models predict performance in important ways - how long individuals take at certain steps is well predicted by the number of subgoals (Anderson et al., 1993)
the difficulty of problems is well predicted by this e.g number of means ends violations (Atwood & Polson, 1976)
how is the PFC implicated in problem solving
closely linked to executive functions planning and goal setting task switching response inhibition working memory
lesions produce characteristic deficits in how problems are solved: rule violation, distractions by other tasks, failure to systematically apply rules from start to finish
how is the MTL implicated in problem solving
Addis et al (2007) propose a role for the MTL in imagining the future
participants were given a scenario to imagine from a first person perspective,
found that there is a widespread area of brain regions that work with the hippocampus for encoding and retrieval
these are not only active when people remember past events but also when they imagine future events
Hassabis et al., 2007
Amnesic patients with damage to the MTL demonstrate deficits in this kind of task
Measured with the richness patients and match controls described some imagined event in the future
On average, patients generate less detailed imaginary events in the future as compared with appropriately matched controls
Hippocampus and surrounding cortex/ connected structures potentially involved in planning
what is the suggested link between spatial and cognitive maps in the (Behrens et al., 2018; Bellmund et al., 2018)
investigated how MTL regions are involved in planning beyond spatial and episodic memory
there is strong evidence for spatial maps
the hippocampus is englarged in taxi drivers and place cells have been found in animal models
the entorhinal cortex has grid cells that fire such that locations are organised as nodes in a grid around triangles in rats, monkeys and humans
there has been the suggestion that the same brain systems may be involved in mapping cognitive domains such as the crockery in your house in terms of colour and shape or vehicles based on properties like power and passenger capacity (Tolmon)
if this is valid it is plausible that we have abstract problem space representations
