Animal Metacognition Flashcards
Why does it matter if animals are metacognitive?
Metacognititive capabilities tied to consciousness, if animals metacognitive suggests conscious in way not characterised as before. Matters in terms of research as can study metacognition with animals in ways can’t in humans, but if metacognitive brings ethical questions of whether we should
Explain the overall arguments about animal behaviour showing metacognition
Animals show the behavioural correlates of uncertainty through information seeking and opting out of difficult decisions. Smith et al. argue this shows metacognition as using representations of uncertainty to guide these behaviours, same as behaviours we see in humans which metacognition is behind so parsimony suggests answer is animals metacognitive too. Le Pelley argues these behaviours can be explained by assoicative learning for which a representation is not neccary, Morgan’s canon suggests simpler answer is better. But associative learning does not explain ancillary behaviours, why some animals capabale of associative learning don’t opt out, or why some animals show metacogntive illusions. Overall seems like animals metacogntivie as more comprehensive explanation than just associative learning.
Explain Call and Carpenter 2001
Experimenter puts food in one of two tubes, subject one attempt to reach in tube for food, information available varied (hidden by screen, delay, control). Animals check/look in the tube before reaching, increases with less infromation/more uncertainty. Animals do this natually/without training, just as human children do. Could be that search is a default object level strategy or using representation of uncertainty.
Explain Smith et al 1995
button get reward if press when high beep, button get reward if press when low beep, third button get smaller garuanteed reward whenever press (via new very easy trial). Dolphins tend to choose opt out button when uncertian/harder to distinguish if high or low, show similar response profile to people. suggested that experiencing uncertainty and use representation of this to choose opt out.
Explain Smith et al. 1997
Opt out behaviour also seen in macaques, reward if press box when has more than certain number of dots, reward if press s button when less dots than this (sparse), third button smaller guaranteed reward whenever press. Humans never told to use third option but report use when uncertain. Suggested that macaque doing same as human. Results replicated in macaque memory, rat decisions, bee decisions etc.
Explain Smith versus Le Pelley
Smith argues that uncertainty behaviours in animals are due to metacognitive processes. Le Pelley argues that uncertainty behaviours are due to associative learning. Opt out choice is seen as a third option chosen when stimulus is unclear/difficult to maximise reward. Simple rienforcement learning models show the same behaviour. Associative learning is simplest explanation so should be taken.
Explain Beran et al. 2009
Capuchin monkeys don’t opt out despite being capabale of assoicative learning genrally and for the paritcular perceptual task - choose middle option (sparse, middle, dense) in three choice condition (large reward if coreect for middle). Associative learning doesn’t explain this, but that other animals using representation of uncertainty could.
Explain ancillary behaviours and give evidence
Behvaiours not part of completing task but see anyway. Dolphins in Smith et al. 1995 showed these by wavering between padels (buttons) and repetitively opening and closing mouth (sign of distress) when task difficult. Allritz et al. 2021 Chimps show certain characteristic behaviours in uncertain situations. Behaviours special to uncertain situations good sign of uncertainty. Associative learning doesn’t explain this as opt out button would just be third option.
Explain Kornell et al. 2007
Perceptual task of judging line length and bet on asnwer (post-decision wagering), monkeys show adpative behaviour betting more when sure of answer/easier trials. Adaptive betting immediately generalised to new perceptual task and across domains to memory task. This generalisation is not explained by task-specific associative learning but can be explained by generalisable metacogntivie representation of uncertainty. Templer et al. 2017 found similar generalisation in rats.
Explain Odegaard et al. 2018
Animals succeptible to positive evidence manipulation as humans are when judging whether more dots moving left or right - monkeys less likely to opt out in high evidence (more dots) versus low evidence (less dots) despite equal difficulty (similar difference in dots moving left versus right). Aligns with people being more confident with more evidence e.g., Ko et al. 2022. Not explained by assoicative learning as is metacogntive illusion, suggests animals using same hueristic cues as humans.
What do neural studies add to the overall argument on animal metacognition?
Orbitofrontal cortex activity aligns with fingerprints of metacognition/acitvity predicted by decision models for metacognitive judgements. Actviity of these cells varies with uncertainty regardles of task domain. Disruption to orbitofrontal cortex via inhibitory injection breaks down adaptive willingness to wait for reward without disrupting task performance. Gives good support for metacognition in animals.
Explain fingerprints in brain and behaviour
Can look for cells with activity that aligns with pattern of acitivty predicted by evidence accumulation based decision models for metacogntive judgements. Characteristic pattern of confidence as function of objective accuracy. As task gets easier and accuracy increases should be more confident in correct responses and less confident in incorrect responses.
Explain Kepecs et al. 2008
Cells in rat orbitofrontal cortex during odour discrimination task (mix of two odours and judge which more of) match fingerprint. Greater activity when more uncertainty (e.g., if more equal mix, if varied delay/time to wait for reward). Aligned with adaptive rat behaviour - more likely to take option to start new trial (pass) when not confident got it right/uncertain.
Explain Lak et al. 2014
Same task set up as Kespecs et al. 2008 (odour discrimination, varied wait for reward). When orbitofrontal cortex shut down via GABA agonsit injection facillitating inhibition, no effect on task performance of odour discrimination but affected waitng behaviour. Normally time willing to wait increases with objective accuracy but breaks down if inhibited. Supports orbitofrontal cortex as substrate for representation of confidence/uncertainty.
Explain Masset et al. 2020
Cells in orbitofrontal cortex that vary with uncertainty/willingness to wait do so regardless of task domain (odour discrimination, auditory task).
