Go when you know: Chimpanzees' confidence movements reflect their responses in a computerized memory task

Flexible information-seeking in chimpanzees

Humans can flexibly use metacognition to monitor their own knowledge and strategically acquire new information when needed. While humans can deploy these skills across a variety of contexts, most evidence for metacognition in animals has focused on simple situations, such as seeking out information about the location of food. Here, we examine the flexibility, breadth, and limits of this skill in chimpanzees. We tested semi-free-ranging chimpanzees on a novel task where they could seek information by standing up to peer into different containers. In Study 1, we tested n = 47 chimpanzees to assess if chimpanzees would spontaneously engage in information-seeking without prior experience, as well as to characterize individual variation in this propensity. We found that many chimpanzees engaged in information-seeking with minimal experience, and that younger chimpanzees and females were more likely to do so. In two subsequent studies, we then further tested chimpanzees who initially showed robust information-seeking on new variations of this task, to disentangle the cognitive processing shaping their behaviors. In Study 2, we examined how a subset of n = 12 chimpanzees applied these skills to seek information about the location versus the identity of rewards, and found that chimpanzees were equally adept at seeking out location and identity information. In Study 3, we examined whether a subset of n = 6 chimpanzees could apply these skills to make more efficacious decisions when faced with uncertainty about reward payoffs. Chimpanzees were able to use information-seeking to resolve risk and choose more optimally when faced with uncertain payoffs, although they often also engaged in information-seeking when it was not strictly necessary. These results identify core features of flexible metacognition that chimpanzees share with humans, as well as constraints that may represent key evolutionary shifts in human cognition.

Metacognition in wild Japanese macaques: cost and stakes influencing information-seeking behavior

Metacognition allows us to evaluate memories and knowledge, thus enabling us to distinguish between what we know and what we do not. Studies have shown that species other than humans may possess similar abilities. However, the number of species tested was limited. Testing ten free-ranging Japanese macaques (Macaca fuscata) on a task in which they had to find food hidden inside one of the four opaque tubes, we investigated whether these subjects would seek information when needed. The monkeys could look inside the tubes before selecting one. We varied three parameters: the baiting process, the cost that monkeys had to pay to look inside the tubes, and the reward at stake. We assessed whether and how these parameters would affect the monkeys' tendency to look inside the tube before selecting one. When they were not shown which tube contained the reward, nine monkeys looked significantly more frequently in at least one condition. Half of them tended to reduce their looks when the cost was high, but only when they already knew the location of the reward. When a high-quality reward was at stake, four monkeys tended to look more inside the tubes, even though they already knew the reward's location. Our results are consistent with those of rhesus macaques, suggesting that metacognitive-like abilities may be shared by Cercopithecidae, and that, at least some monkeys may be aware of their lack of knowledge.

State-transition-free reinforcement learning in chimpanzees (Pan troglodytes)

The outcome of an action often occurs after a delay. One solution for learning appropriate actions from delayed outcomes is to rely on a chain of state transitions. Another solution, which does not rest on state transitions, is to use an eligibility trace (ET) that directly bridges a current outcome and multiple past actions via transient memories. Previous studies revealed that humans (Homo sapiens) learned appropriate actions in a behavioral task in which solutions based on the ET were effective but transition-based solutions were ineffective. This suggests that ET may be used in human learning systems. However, no studies have examined nonhuman animals with an equivalent behavioral task. We designed a task for nonhuman animals following a previous human study. In each trial, participants chose one of two stimuli that were randomly selected from three stimulus types: a stimulus associated with a food reward delivered immediately, a stimulus associated with a reward delivered after a few trials, and a stimulus associated with no reward. The presented stimuli did not vary according to the participants' choices. To maximize the total reward, participants had to learn the value of the stimulus associated with a delayed reward. Five chimpanzees (Pan troglodytes) performed the task using a touchscreen. Two chimpanzees were able to learn successfully, indicating that learning mechanisms that do not depend on state transitions were involved in the learning processes. The current study extends previous ET research by proposing a behavioral task and providing empirical data from chimpanzees.

Don't look back on failure: spontaneous uncertainty monitoring in chimpanzees

During computer-controlled cognitive tasks, chimpanzees often look up at the food dispenser, which activates at the same time as feedback for the correct choice but not for feedback for the incorrect choice. Do these "looking back" behaviors also indicate signs of spontaneous monitoring of their confidence in their choices? To address this question, we delayed the feedback for 1 s after their choice responses and observed their look-back behaviors during the delay period. Two chimpanzees looked up at the food dispenser significantly less frequently when their choice was incorrect (but the feedback was not given) than when it was correct. These look-back behaviors have not been explicitly trained under experimental contexts. Therefore, these results indicate that chimpanzees spontaneously change the frequency of their look-back behaviors in response to the correctness or incorrectness of their own choices, even without external feedback, suggesting that their look-back behaviors may reflect the level of "confidence" or "uncertainty" of their responses immediately before.

Information seeking in western lowland gorillas (Gorilla gorilla gorilla)

Many animals will seek information when they do not know the answer to a problem, suggesting that they monitor their knowledge state. In the classic "tubes task," subjects are presented with a set of opaque tubes and either see (visible trials) or do not see (hidden trials) which tube holds a food reward on a given trial. Chimpanzees, bonobos, orangutans, and human children show information-seeking behavior on this task, looking into the tubes to find the reward more on hidden than on visible trials. However, evidence for information-seeking behavior in gorillas is limited. In Experiment 1, gorillas that were presented with a classic information-seeking tubes task showed performance patterns consistent with metacognitive behavior; they looked down tubes more on hidden than on visible trials, their accuracy on hidden trials on which they looked was higher than on hidden trials on which they did not look, and they primarily employed an appropriate search strategy when looking down the tubes. In Experiment 2, we decreased or increased the amount of effort required to look down the tubes by increasing or decreasing the height of the tubes, respectively. Gorillas were less likely to look in tubes on trials that required high effort, but continued to look more on hidden than on visible trials, indicating that their tendency to look was affected by both knowledge state and effort. Together these results provide strong evidence for logical, controlled information-seeking behavior by gorillas.

Social cognition and metacognition in great apes: a theory

Twenty-five years ago, at the founding of this journal, there existed only a few conflicting findings about great apes' social-cognitive skills (theory of mind). In the 2 ½ decades since, we have discovered that great apes understand the goals, intentions, perceptions, and knowledge of others, and they use this knowledge to their advantage in competitive interactions. Twenty-five years ago there existed basically no studies on great apes' metacognitive skills. In the 2 ½ decades since, we have discovered that great apes monitor their uncertainty and base their decisions on that, or else decide to gather more information to make better decisions. The current paper reviews the past 25 years of research on great ape social cognition and metacognition and proposes a theory about how the two are evolutionarily related.

Chimpanzees (Pan troglodytes) show subtle signs of uncertainty when choices are more difficult

Humans can tell when they find a task difficult. Subtle uncertainty behaviors like changes in motor speed and muscle tension precede and affect these experiences. Theories of animal metacognition likewise stress the importance of endogenous signals of uncertainty as cues that motivate metacognitive behaviors. However, while researchers have investigated second-order behaviors like information seeking and declining difficult trials in nonhuman animals, they have devoted little attention to the behaviors that express the cognitive conflict that gives rise to such behaviors in the first place. Here we explored whether three chimpanzees would, like humans, show hand wavering more when faced with more difficult choices in a touch screen transitive inference task. While accuracy was very high across all conditions, all chimpanzees wavered more frequently in trials that were objectively more difficult, demonstrating a signature behavior which accompanies experiences of difficulty in humans. This lends plausibility to the idea that feelings of uncertainty, like other emotions, can be studied in nonhuman animals. We propose to routinely assess uncertainty behaviors to inform models of procedural metacognition in nonhuman animals.

Outcome expectancy and suboptimal risky choice in nonhuman primates

Animals will favor a risky option when a stimulus signaling reward bridges the choice and the outcome. The present experiments investigated signal-induced risky choices and reward-outcome expectations in rhesus and capuchin monkeys. Risky choice was assessed by preference for a large-probabilistic reward over a modest-certain reward. Outcome expectancy was assessed by providing a truncation-response to shorten the delay period. In Experiment 1 both species generally favored the risky option compared to a safe option when the outcomes were signaled and generally shortened the delays except when a signaled-loss stimulus was presented. The use of the delay-truncation response suggested that the monkeys were sensitive to the information conveyed by the stimulus. Experiments 2 and 3 were designed to investigate whether the delay-truncation response used by capuchin monkeys was strategically used reflecting explicit decision-making versus a conditioned response to reward stimuli. A perceptual judgment task was included and the selective use of the delay-truncation response on unsignaled correct trials may suggest the involvement of metacognitive processes. The capuchin monkeys generally truncated the delays except under conditions where reward would not be expected (risky-loss or incorrect-judgment). When the outcomes were unsignaled during the delay some capuchin monkeys were less likely to truncate the delay following an incorrect task response. Overall, the monkeys: (1) made more risky choices when the outcomes were signaled - consistent with gambling-like behavior. (2) selectively truncated the unsignaled delays when rewards could be anticipated (even when metacognitive-like awareness guided anticipation) - suggesting that delay truncation responses reflect explicit outcome expectancy.

Investigating information seeking in ravens (Corvus corax)

Measuring the responses of non-human animals to situations of uncertainty is thought to shed light on an animal’s metacognitive processes; namely, whether they monitor their own knowledge states. For example, when presented with a foraging task, great apes and macaques selectively seek information about the location of a food item when they have not seen where it was hidden, compared to when they have. We presented this same information seeking task to ravens, in which a food item was hidden in one of three containers, and subjects could either watch where the food was hidden, infer its location through visual or auditory clues, or were given no information. We found that unlike several ape species and macaques, but similar to capuchin monkeys, the ravens looked inside at least one tube on every trial, but typically only once, inside the baited tube, when they had either witnessed it being baited or could visually infer the reward’s location. In contrast, subjects looked more often within trials in which they had not witnessed the baiting or were provided with auditory cues about the reward’s location. Several potential explanations for these ceiling levels of looking are discussed, including how it may relate to the uncertainty faced by ravens when retrieving food caches.

Capuchin monkeys (sometimes) go when they know: Confidence movements in Sapajus apella

To test for evidence of metacognition in capuchin monkeys (Sapajus apella), we analyzed confidence movements using a paradigm adapted from research with chimpanzees. Capuchin monkeys provide an interesting model species for the comparative assessment of metacognition as they show limited evidence of such cognitive-monitoring processes in a variety of metacognition paradigms. Here, monkeys were presented with a computerized delayed matching-to-sample (DMTS) memory test in one location but were rewarded for correct responses in a separate location. Movements could be made from one location to the other at any time, but movements between a response and reward feedback may reflect confidence in the accuracy of the response. Critically, DMTS tests included occasional "no sample" trials where monkeys' performance was at chance when the trial started without a sample and a 1-s interval to the response options. We predicted that monkeys would (1) perform less accurately (and less confidently) at longer retention intervals, (2) move to the dispenser early more often on trials completed correctly than incorrectly, and (3) show a relation between faster response latency and early movements. Analyses of response times and "go" or "no go" confidence movements before feedback to the reward location suggested that the monkeys were capable of monitoring confidence in their responses. However, their confidence movements were less precise and less flexible than chimpanzees. Overall, this paradigm can reveal potential metacognitive abilities in nonhuman animals that otherwise demonstrate these abilities inconsistently.

Developing a Reflective Mind: From Core Metacognition to Explicit Self-Reflection

Metacognition is the ability to monitor and control cognition. Because young children often provide inaccurate metacognitive judgments when prompted to do so verbally, it has long been assumed that this ability does not develop until late childhood. This claim is now challenged by new studies using nonverbal paradigms and revealing that basic forms of metacognition—such as the ability to estimate decision confidence or to monitor errors—are present even in preverbal infants. This new line of evidence suggests that young children adapt to their environment not only by considering their physical and social surroundings but also by reflecting on their own cognitive states.

Simultaneous versus prospective/retrospective uncertainty monitoring: The effect of response competition across cognitive levels

Early animal-metacognition researchers singled out simultaneous metacognition paradigms for theoretical criticism, because these paradigms presented concretely rewarded perceptual responses and the metacognitive response simultaneously. This method potentially introduced associative cues into the situation that could confound the interpretation of the metacognitive response. Evaluating this possibility, we compared humans' metacognitive performances in simultaneous and nonsimultaneous (prospective, retrospective) paradigms that were otherwise identical. Results show that the metacognition response in these tasks is not prompted by associative cues arising from the simultaneous task format. To the contrary, the metacognitive response is used more robustly and accurately when it is removed from direct competition with the primary perceptual responses. Thus, early researchers were correct to judge that the nonsimultaneous paradigms tap metacognition more robustly and sensitively. However, this is probably true because the simultaneous paradigm mingles responses adjudicated on two different cognitive-processing levels. And, in that case, the metacognitive response can be outcompeted and suppressed by the salient presence of primary, concretely rewarded perceptual responses. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Social Learning Strategies: Bridge-Building between Fields

While social learning is widespread, indiscriminate copying of others is rarely beneficial. Theory suggests that individuals should be selective in what, when, and whom they copy, by following 'social learning strategies' (SLSs). The SLS concept has stimulated extensive experimental work, integrated theory, and empirical findings, and created impetus to the social learning and cultural evolution fields. However, the SLS concept needs updating to accommodate recent findings that individuals switch between strategies flexibly, that multiple strategies are deployed simultaneously, and that there is no one-to-one correspondence between psychological heuristics deployed and resulting population-level patterns. The field would also benefit from the simultaneous study of mechanism and function. SLSs provide a useful vehicle for bridge-building between cognitive psychology, neuroscience, and evolutionary biology.

Chimpanzees show some evidence of selectively acquiring information by using tools, making inferences, and evaluating possible outcomes

Metacognition refers to thinking about one's thinking or knowing what one knows. Research suggests that this ability is not unique to humans and may be shared with nonhuman animals. In particular, great apes have shown behaviors on a variety of tasks that are suggestive of metacognitive ability. Here we combine a metacognitive task, the information-seeking task, with tool use and variable forms of initial information provided to chimpanzees to explore how informational states impact behavioral responses in these apes. Three chimpanzees were presented with an apparatus that contained five locations where food could be hidden. If they pointed to the correct location, they received the reward, but otherwise they did not. We first replicated several existing findings using this method, and then tested novel hypotheses. The chimpanzees were given different types of information across the experiments. Sometimes, they were shown the location of the food reward. Other times, they were shown only one empty location, which was not useful information. The chimpanzees also could use a tool to search any of those locations before making a selection. Chimpanzees typically used the tool to search out the location of the reward when they could not already know where it was, but they did not use the tool when they already had been given that information. One chimpanzee made inferences about the location of hidden food, even when that food was never shown in that location. The final experiment involved hiding foods of differing preference values, and then presenting the chimpanzees with different initial knowledge states (i.e., where the best food was located, where the less-preferred food was located, or where no food was located). All chimpanzees used the tool when they needed to use it to find the best possible item on that trial, but responded by choosing a location immediately when they did not need the tool. This finding highlights that their behavior was not the result of a simple rule following such as pointing to where any food had been seen.

Gambling in rhesus macaques (Macaca mulatta): The effect of cues signaling risky choice outcomes

Preference for a larger-variable "risky" option over a smaller-reliable "safe" option often depends upon the likelihood that the risky option will deliver a sufficiently sized reward to have an equivalent or superior expected value. However, preference for the risky option has been shown to increase under conditions where informative stimuli signaling the outcome of a risky choice is included between the choice and the outcome and this risk-prone preference persists even when the risky option has a lower expected value than the alternative safe option. In the present study, rhesus macaques chose between a risky option and a safe option across two experimental phases to determine whether the outcome signal affected the degree of preference for the risky option. Overall, six out of seven macaques showed a greater preference for the risky option in the signaled condition than in the unsignaled condition. The macaques' risky choices were sensitive to the expected value of the risky option and the signaled condition produced a general increase in risky choices independently of the expected value of the risky outcome. Overall, these results are consistent with those obtained with other animals, and this may relate to a process where animals show a biased preference for "good news." This process may model some of the relevant factors that explain the psychology of gambling in humans.

Primate cognition: attention, episodic memory, prospective memory, self-control, and metacognition as examples of cognitive control in nonhuman primates

Primate Cognition is the study of cognitive processes, which represent internal mental processes involved in discriminations, decisions, and behaviors of humans and other primate species. Cognitive control involves executive and regulatory processes that allocate attention, manipulate and evaluate available information (and, when necessary, seek additional information), remember past experiences to plan future behaviors, and deal with distraction and impulsivity when they are threats to goal achievement. Areas of research that relate to cognitive control as it is assessed across species include executive attention, episodic memory, prospective memory, metacognition, and self-control. Executive attention refers to the ability to control what sensory stimuli one attends to and how one regulates responses to those stimuli, especially in cases of conflict. Episodic memory refers to memory for personally experienced, autobiographical events. Prospective memory refers to the formation and implementation of future-intended actions, such as remembering what needs to be done later. Metacognition consists of control and monitoring processes that allow individuals to assess what information they have and what information they still need, and then if necessary to seek information. Self-control is a regulatory process whereby individuals forego more immediate or easier to obtain rewards for more delayed or harder to obtain rewards that are objectively more valuable. The behavioral complexity shown by nonhuman primates when given tests to assess these capacities indicates psychological continuities with human cognitive control capacities. However, more research is needed to clarify the proper interpretation of these behaviors with regard to possible cognitive constructs that may underlie such behaviors. WIREs Cogn Sci 2016, 7:294-316. doi: 10.1002/wcs.1397 For further resources related to this article, please visit the WIREs website.

Who Knows? Metacognitive Social Learning Strategies

To make good use of learning from others (social learning), we need to learn from the right others; from agents who know better than we do. Research on social learning strategies (SLSs) has identified rules that focus social learning on the right agents, and has shown that the behaviour of many animals conforms to these rules. However, it has not asked what the rules are made of, that is, about the cognitive processes implementing SLSs. Here, I suggest that most SLSs depend on domain-general, sensorimotor processes. However, some SLSs have the characteristics tacitly ascribed to all of them. These metacognitive SLSs represent 'who knows' in a conscious, reportable way, and have the power to promote cultural evolution.