Download e-book Social Learning: An Introduction to Mechanisms, Methods, and Models

Free download. Book file PDF easily for everyone and every device. You can download and read online Social Learning: An Introduction to Mechanisms, Methods, and Models file PDF Book only if you are registered here. And also you can download or read online all Book PDF file that related with Social Learning: An Introduction to Mechanisms, Methods, and Models book. Happy reading Social Learning: An Introduction to Mechanisms, Methods, and Models Bookeveryone. Download file Free Book PDF Social Learning: An Introduction to Mechanisms, Methods, and Models at Complete PDF Library. This Book have some digital formats such us :paperbook, ebook, kindle, epub, fb2 and another formats. Here is The CompletePDF Book Library. It's free to register here to get Book file PDF Social Learning: An Introduction to Mechanisms, Methods, and Models Pocket Guide.
Scientists refer to this as social learning. It is one of the most exci Social Learning: An Introduction to Mechanisms, Methods, and Models. William Hoppitt.
Table of contents

In this event-based setup, actions of foragers can overlap in time, and some foragers can complete multiple quick actions e. The learning algorithms include representations of individual and social learning, and update long term memory about properties of resources that foragers interact with as a consequence of their decisions. Life-history updating occurs at regular time intervals and includes: i metabolism or energy expenditure; ii digestion of consumed resources; iii deaths and iv births of foragers; and v splitting of groups.

After a forager dies, a forager is selected from the remaining population to reproduce, thus maintaining a fixed population size. Foragers are selected to reproduce in relation to their energy levels, where a doubling in energy leads to an 8-fold increase in the probability to reproduce. In simulations with grouping, groups grow due to births until they reach a maximum size, and then split randomly into two equally sized daughter groups.

Groups shrink due to deaths and disappear when the last group member dies. Spatio-temporal scaling: The environment is a continuous space of about 40 km 2 , foragers take steps of a meter at a speed of 0. Foragers can observe resources up to 2 meters away, and can observe which resources their neighbors are interacting with at 20 meters a best case scenario for social learning, Additional file 3. There are no constraints on observing group members for grouping purposes in order to ensure cohesive groups, but the spread of groups tends to be in the order of 5—40 meters.

All movement occurs in continuous space and there are no constraints on direction. Additional file 3: Video showing group foragers observing each other, which relevant for both stimulus enhancement and observational learning. When a forager observes another forager the foragers are connected by an olive-green line. MP4 kb 2. In the model a year is defined as days, and a day is 12 hours or minutes, where we focus on daylight time in a day.

Social Learning: An Introduction to Mechanisms, Methods, and Models (eBook, ) []

Thus foragers can complete many hundreds of behavioral actions in a day and learn from them. Energy expenditure metabolism occurs every minute. Foragers can live maximally for 20 years, but can die before that at any minute. In our default setting, resource items of resource types are distributed in patches with items each.

Login using

There are 50 patch types, and a patch type is characterized by the presence of five resources types that only occur in that patch type as in trees with fruit, leaves, flowers etc. In order to generate variation across patches of a given type, each patch of a given type is defined by three resource types which are randomly selected from the five resource types that characterize that patch type. While these parameter values typically underestimate the diversity of natural environments, we strike a pragmatic balance between model complexity and simulation environments that are too simple, and where learning hardly plays a role [ 24 ].

We compare this ecological context with randomly distributed resources without patches, and pure patches where each patch type has only one resource type.

Social learning - a framework

Resource items disappear when consumed by foragers, and are then unavailable for consumption. Environmental change occurs randomly at any minute with a given probability and changes a randomly selected resource type into another newly generated resource type which is unfamiliar to the foragers. For ease of interpretation we express this as a rate, namely how many resource types change per year EC.

Original Research ARTICLE

All resource items of the disappearing type change into the new resource type. We vary environmental change EC across simulations to determine the effect of environmental change. We compare this kind of environmental change to one where resources do not disappear and change into new ones, but where resources remain familiar but change in quality. The quality of a resource type Q r is drawn from a random distribution with mean 0. Thus we generate variation in quality across resource types which enables the learning process to be studied as an optimization process.

Quality defines the maximal reward that a forager can obtain from a resource type when it has sufficient experience with processing that resource type. S r varies randomly between 1 and 4 integer values only and H r is varied across simulations to determine an overall difficulty of learning in the environment. These observations are relevant for stimulus enhancement SE and observational learning OL. This attraction-alignment algorithm ensures that foragers stay together but travel in a relatively efficient manner through the environment. The 20 items are randomly selected from those in view.

The search terminates as soon as an item is chosen for consumption, or when none of the items is chosen. During evaluation of a resource item, these three factors come together to determine the probability P F to choose to eat that item as follows:. For solitary foragers this means that the exploration rate P E must be greater than zero. For grouping foragers, social stimulation P S could in principle replace exploration P E as the means to sample unknown resources. Selectivity is adjusted relative to environmental conditions by adjusting the expected quality of the environment a ie Fig.

Each time the forager is too selective, it does not fill its stomach and reduces its selectivity, and vice versa. As a result, a ie is tuned in order to optimise energy intake, within the constraints of the algorithm. Qualitatively, this selection algorithm can give rise to the optimal food choice rule [ 27 ] where only resources above a certain perceived quality are eaten and all others are ignored zero-one rule.

Note however that our algorithm works on perceived quality and not actual quality since the foragers are learning about resource quality and are not omniscient. Satiation aversion: foragers develop temporary aversions after becoming satiated stomach filled with a given resource type. This model specification was added to ensure that foragers consume a diverse set of resource types [ 21 ], as is typical for primates and as was assumed in previous models [ 21 , 24 ]. In the absence of any social influences on learning, learning in our model is composed of i exploration, ii skill learning, and iii reinforcement learning about rewards associated with resources.

Navigation menu

To enable foragers to sample partially unfamiliar resource types, and hence to start learning, we implemented exploration. After processing resource items, foragers develop skill, which increases the rewards they can obtain from resources items of that type. After consuming resource items, foragers develop expectations about rewards via reinforcement, and can use those to decide what to eat.

We do not expect this to affect the results qualitatively. Exploration: The probability that a forager explores an item of resource type r is:.

  • Jesus the Temple!
  • Social learning theory;
  • Co-Evolution of Social Learning and Evolutionary Preparedness in Dangerous Environments.
  • The Accommodated Animal: Cosmopolity in Shakespearean Locales.
  • Social Learning: An Introduction to Mechanisms, Methods, and Models;
  • About De Gruyter.

Certainty was included to ensure that foragers do not continue exploring when already highly familiar with resources. However, when rewards from resource types no longer change, for instance because skill levels are high, certainty becomes high, and foragers end up with a low tendency to explore that resource type. Certainty c ir is updated as follows:. Experience t ir is the total time a forager i has spent processing a resource type r in its life, and increases each time the forager processes and consumes a resource item of type r.

Skill s ir determines the reward e ir forager i obtains from resource type r as a function of resource quality Q r :. Resource types with high H Fig. Note that for simplicity we assume that while for different foragers the same resource items can provide different energy, depletion from the environment and the number of items that can be eaten is the same.

This can be interpreted as foragers consuming a certain amount of resource in a given amount of time irrespective of how well it is processed, but that energy obtained depends on processing.

Associated Data

Moreover the item is then no longer available for other foragers. Reinforcement learning about expected rewards: The rewards that foragers associate with each resource type r are updated via reinforcement as follows:. This corresponds to a Rescorla-Wagner model [ 30 ] where all stimuli have the same salience. Associations are initially non-existent i. Local enhancement LE : Arises spontaneously through grouping behaviour, since individuals are inclined to approach locations in which other members of their group are found, and thereafter to interact with resources in those regions.

We therefore do not directly implement local enhancement, but it emerges spontaneously as soon as foragers move in groups [ 24 ].

The local enhancement that we consider is coarse grained, and does not direct individuals to particular resources, or to features of those resources. The impact of the demonstrator depends on the social learning mechanism. Only one resource type r is subject to SE at a time.

Social Learning An Introduction To Mechanisms Methods And Models

SE does not directly affect expected rewards or skill. Greater observation time leads to greater skill acquisition, where maximal observation time is the maximal time it takes to process and consume a resource M. Observation does not provide information about rewards. Energy accumulates if energy intake from food exceeds metabolism and reproduction costs.

A limited stomach capacity and digestion intervals were added to the model to ensure selective foraging, as is typical for primates and as was assumed in previous models [ 21 , 24 ]. In addition, an explicit metabolism cost, ensures that there is a viability constraint in the model, where foragers must gain enough energy from food otherwise they die.

Foragers die of old age at 20 years , stochastically determined deaths, or starvation. Births occur as a function of energy reserves each time a forager dies, keeping the population constant at size N , where probability that forager i reproduces is:. Parameter combinations that lead to greater energy levels lead to faster rates of reproduction.

Thus parameters can vary between individuals and can evolve over time via inheritance to offspring, mutation and natural selection. The mutation rate was selected operationally such that parameters evolve consistently within a reasonable time frame. There is no migration between groups.