Categories and Inductive Inferences

JoVE 科学教育
Developmental Psychology
需要订阅 JoVE 才能查看此.  登录或开始免费试用。
JoVE 科学教育 Developmental Psychology
Categories and Inductive Inferences

5,097 Views

10:08 min

April 30, 2023

概述

Source: Laboratories of Nicholaus Noles and Judith Danovitch—University of Louisville

It might be possible for the human brain to keep track of each individual person, place, or thing encountered, but that would be a very inefficient use of time and cognitive resources. Instead, humans develop categories. Categories are mental representations of real things that can be used for a variety of purposes. For example, individuals can use the perceptual features of animals to place them into a given category. So, upon seeing a furry, four-legged, tail-wagging, barking animal, a person can determine that it is a dog. This is one of many examples where people use perceptual similarity to fit new experiences into their existing mental representations.

However, category membership is much more than skin deep, especially for representations of animals. Frank Keil demonstrated this by using a simple, yet powerful technique that focused on the differences between natural kinds and artifacts. Natural kinds include animals and other living things, while artifacts consist largely of nonliving things, such as tables or gold bricks. In his study, Keil told children stories about natural kinds and artifacts that underwent transformations causing them to cross categorical boundaries. For example, he described a step-by-step process by which a raccoon was transformed into a creature that resembled a skunk in every way. At the end of the story, the raccoon was black with a white stripe, and it had implanted glands that made it smell like a skunk, too. He asked the children to determine whether the resulting animal was a raccoon or a skunk. He used a similar method to describe the transformation of a tire-an artifact-into a shoe. Children's responses revealed interesting developmental changes into how people think about artifacts and natural kinds.

This video demonstrates Frank Keil's transformation study.1

Procedure

Recruit healthy 5-, 7-, and 9-year-old children with normal hearing and vision and no history of developmental disorders. For the purposes of this demonstration, only one child is tested. Larger sample sizes (as in Frank Keil's transformation study1) are recommended when conducting any experiments.

1. Data collection

  1. Gather the necessary materials.
    1. Design eight vignettes describing the transformation of animals from one kind to another via surgery.
      1. Natural kind vignette example: "The doctors took a raccoon and shaved away some of its fur. They dyed the rest of the fur black. Then, they bleached a single white stripe down the center of its back. They surgically added a sack of super smelly stuff into its body, just like a skunk has. When they were all done, the animal looked like this. After the operation, was this a skunk or a raccoon?"
    2. Design eight vignettes describing the transformation of artifacts from one kind to another via physical alteration.
      1. Artifact vignette example: "The doctors took a coffeepot that looked like this. They sawed off the handle, sealed the top, took off the top knob, sealed the spout, and sawed it off. They also sawed off the base and attached a flat piece of metal. They attached a little stick, cut a window in it, and filled the metal container with bird food. When they were done, it looked like this. After the operation, was this a coffeepot or a birdfeeder?"
    3. Create cards with images of each item before and after transformation.
  2. Preamble
    1. Present the child with a preamble introducing doctors and describing operations and surgery.
  3. Test
    1. Present the child with each of the 16 vignettes in a random order.
    2. If the child tries to give an ambiguous response (e.g., "It's both" or "It's a skuncoon"), ask them to pick one of the two highlighted responses.
    3. Question the child in a free-form manner in order to understand what aspects of the vignette guided their responses. For example, the experimenter might ask questions like: "Why is it a birdfeeder?" or "If the animal looks like a skunk and smells like a skunk, why did you say it was a raccoon?" This questioning helps to determine what features or principles the child uses to guide their intuitions.
    4. Record the child's responses, and then transcribe them for future analysis.

2. Analysis

  1. Two independent coders read the child's transcriptions and code them on a scale from 1 to 3 for each item. This score is an index of the child's resistance to categorical change.
    1. A score of 1 on an item means a child shifts their intuitions to follow the transformation operation. These responses focus on characteristic features.
      1. Example: "It's stinky and stripy like a skunk, so it's a skunk."
    2. A score of 2 on an item means a child's responses are indecisive, indicating that they are unsure if the transformation results in a change in the item's kind.
      1. Example: "It's a rac-unk or a skoon. I don't know."
    3. A score of 3 on an item means that a child resists shifting their intuitions following transformation operations. These responses contrast with observed features.
      1. Example: "It doesn't matter what it looks like on the outside; it's still a raccoon."
  2. Use an analysis of variance to determine if there are differences between the three age groups and two types of items presented to the children.

To organize the world and efficiently use cognitive resources, children learn to place objects, people, and locations into categories.

In a broad sense, a category—like fruits—is a mental representation of a collection of real things that share similar features, such as the presence of seeds.

In this instance, the characteristic of having seeds is a “hidden” or internal feature—one not easily observed—that can be used to link items to form a category.

However, sometimes the feature—or features—on which a category is based can be readily perceived. For example, if a child sees an animal that has four legs, a tail, and barks, they’ll likely categorize it as a dog.

Importantly, children can take what they know about a single item belonging to a category, like their pet dog, and use that information to identify and make inductive inferences—educated guesses—about unfamiliar category members.

For example, if a child knows that their dog wags its tail when it’s happy, then they can guess that a new tail-wagging dog is also happy.

Based on the pioneering work of Frank Keil, this video explores how to investigate the criteria 5- to 9-year-olds use to categorize objects by designing stimuli that cross categorical boundaries, and explains how to perform such a transformational study and interpret categorization data. We will also look at how psychologists use this technique in other applications, for example to explore how children make inferences.

In this experiment, researchers present 5-, 7-, and 9-year-old children with 16 stories in which items are physically altered so that they resemble something else.

Eight of these stories center around living animals or naturally occurring objects, like zebras or diamonds—collectively called natural kinds—and eight deal with man-made, non-living items referred to as artifacts, such as cups or pots.

For instance, in a natural kind-focused transformation story, children are first shown a “before” image of a naturally occurring item or animal, such as a raccoon.

They’re then told that doctors surgically transformed this thing, for example by attaching containers of foul-smelling liquid to the raccoon, and shaving and dying portions of its fur so that it becomes black and white with a puffy tail.

Importantly, the researcher emphasizes that such operations result in the item being transformed so that it represents a different object of the same type: a living natural kind is made to look like another living natural kind.

Children are finally presented with an “after” image of the subject, and asked how they’d categorize it post-surgery—if it’s a raccoon or a skunk.

Their answers are rated on a scale of 1 to 3. In this system, if children shift their categorization—if they assert that the shaved raccoon is now a skunk, as it resembles this animal in appearance and smell—they’re given a score of 1.

In contrast, if children are unsure of how the transformed thing should be categorized—if it’s some “ra-kunk” hybrid of the two—they’re given a score of 2.

Finally, if children don’t change the initial categorization of the item—saying that the animal remains a raccoon, despite its altered physical state—they’re assigned a score of 3.

Here, the dependent variables are the average scores children receive across natural kind and artifact stories, which can be used to gauge children’s tendency to resist changing their identification of transformed objects.

Based on research performed by Frank Keil and colleagues, it is expected that children will readily change their categorization of inanimate artifacts, but not of living natural kinds, presumably due to some internal, unchangeable feature that animals possess.

Before the experiment begins, create cards with images of pre- and post-transformation items. Design 16 such transformation vignettes—eight dealing with artifacts, and eight dealing with natural kinds.

When the child arrives, greet them and introduce them to the concepts of doctors and surgeries.

Present the 16 transformation vignettes to them in a random order: “The doctors took a coffeepot that looked like this. They sawed off the handle, sealed the top, took off the top knob, sealed the spout, and sawed it off. They also sawed off the base and attached a flat piece of metal. They attached a little stick, cut a window in it, and filled the metal container with bird food. When they were done, it looked like this.”

“After the operation, was this a coffeepot or a birdfeeder?”

If in any instance the child gives an ambiguous or hybrid response—like saying that a coffeepot transformed into a birdfeeder is now a “coffee-feeder”—also ask them to call it by either its pre- or post-transformation name. This encourages children to think about their answers, as well as warrants different responses across all of the vignettes.

Question the child in a free-form manner about the reasons behind their response, in order to determine the features or principles of the post-transformation item that led them to their conclusion.

For each vignette, record the child’s response for future analysis.

When the child has completed all 16 vignettes, have two independent raters read their responses and code them on a scale of 1–3.

To analyze the data for each of the two types of transformations—either natural kinds or artifacts—plot the average children’s score as a function of age.

Perform an analysis of variance to determine if there are differences between the three age groups or two types of transformations.

Notice that as age increases, the average scores of children for the artifact transformations remained relatively constant, hovering around 1.25. This indicates that children at all ages studied readily change their categorization of altered artifacts.

In addition, these results suggest that the perceptual features of post-transformation items in these instances—what a child sees—informs their categorical placement. In contrast, for natural kind transformations—especially those involving animals—children’s average scores increased as a function of age. This indicates that as children get older, they gradually represent category membership as an internal, unchangeable aspect of animals.

Now that you know how researchers are using transformation scenarios to better understand how children categorize general artifacts and natural kinds, let’s look at other applications of this technique.

As you’ve learned, natural kinds can be either living items—like plants or animals—or inanimate objects, like minerals.

Thus, some researchers are further looking into how children come to categorize these different types of natural kinds, and if any differences exist between these processes.

Such work has demonstrated that whereas children quickly learn to categorize animate natural kinds—and resist the tendency to change their categorization of animals post-transformation—a similar ability for inanimate natural kinds develops more slowly.

Some psychologists have suggested that parental influence may play a role in this phenomenon. For example, a parent may emphasize that an animal should be categorized as a fish if its parents were fish, and if it lays eggs that also yield fish.

However, given the advanced subject matter of what makes an inanimate natural kind—like the organization of electrons in its atoms, or the elements that constitute it—parents may not have similar discussions with their children about such items.

Other researchers are looking at how children apply facts about a single member of a category to make inferences about other items in the same category.

For example, psychologists could tell children that a mouse is more active at night and likes to eat cheese, and then show them a collection of realistic pictures composed of mice with different coats, other rodents, and unrelated animals—like a cow.

By either including or removing labels from these images, researchers can evaluate the extent to which textual category labels, in addition to morphological features, factor in children’s ability to infer that other mice have the same characteristics.

You’ve just watched JoVE’s video on how children develop the ability to categorize—and make inferences about—natural kinds and artifacts. By now, you should know how to design and use transformation stories to investigate this phenomenon, and collect and interpret results. You should also have an understanding of how psychologists are using this method to investigate other aspects of the categorization process.

Thanks for watching!

Results

In order to have enough power to see significant results, researchers would have to test at least 18 children in each age group. Typically, when asked about artifacts, children in all three age groups conclude that what is seen confirms the categorical placement. If a tire is transformed into a rubber shoe, then it is a shoe and not a tire. In contrast, children presented with natural kinds reveal a developmental trend. Five-year-olds are either indecisive or see an animal’s posttransformation features as indicating their category membership. As children get older, they determine increasingly often that animals remain the same kind of thing in spite of any physical transformation they may undergo. This experiment demonstrates that children represent category membership as an internal, unchangeable aspect of animals increasingly as they get older, and this idea drives children’s intuitions about category membership (Figure 1).

Figure 1
Figure 1. The average tendency for children to resist shifts in category membership. Low numbers indicate that transforming a target’s features changes its category membership.

Applications and Summary

Frank Keil's work demonstrates that internal characteristics count. Children treat category membership as springing from internal characteristics that cause animals' appearances and behaviors, and children continue to have the intuition that animals belong to their category, even when appearances and behaviors change. Generally, this finding supports other work demonstrating that children use categorical information and not other cues, such as appearance, to guide their inferences about animals. For example, individuals can use categories to make inductive inferences, or educated guesses, based on their categorical knowledge. So, if a child knows it's dangerous to touch their pet cat when its tail is wagging, then they can make an inductive inference that any new, tail-wagging cat is also dangerous. These inferences, like the inferences of category membership in Keil's study, are driven by category membership and not necessarily appearance.

References

  1. Keil, F.C. Concepts, Kinds, and Cognitive Development. MIT Press: Cambridge, MA (1989).

成績單

To organize the world and efficiently use cognitive resources, children learn to place objects, people, and locations into categories.

In a broad sense, a category—like fruits—is a mental representation of a collection of real things that share similar features, such as the presence of seeds.

In this instance, the characteristic of having seeds is a “hidden” or internal feature—one not easily observed—that can be used to link items to form a category.

However, sometimes the feature—or features—on which a category is based can be readily perceived. For example, if a child sees an animal that has four legs, a tail, and barks, they’ll likely categorize it as a dog.

Importantly, children can take what they know about a single item belonging to a category, like their pet dog, and use that information to identify and make inductive inferences—educated guesses—about unfamiliar category members.

For example, if a child knows that their dog wags its tail when it’s happy, then they can guess that a new tail-wagging dog is also happy.

Based on the pioneering work of Frank Keil, this video explores how to investigate the criteria 5- to 9-year-olds use to categorize objects by designing stimuli that cross categorical boundaries, and explains how to perform such a transformational study and interpret categorization data. We will also look at how psychologists use this technique in other applications, for example to explore how children make inferences.

In this experiment, researchers present 5-, 7-, and 9-year-old children with 16 stories in which items are physically altered so that they resemble something else.

Eight of these stories center around living animals or naturally occurring objects, like zebras or diamonds—collectively called natural kinds—and eight deal with man-made, non-living items referred to as artifacts, such as cups or pots.

For instance, in a natural kind-focused transformation story, children are first shown a “before” image of a naturally occurring item or animal, such as a raccoon.

They’re then told that doctors surgically transformed this thing, for example by attaching containers of foul-smelling liquid to the raccoon, and shaving and dying portions of its fur so that it becomes black and white with a puffy tail.

Importantly, the researcher emphasizes that such operations result in the item being transformed so that it represents a different object of the same type: a living natural kind is made to look like another living natural kind.

Children are finally presented with an “after” image of the subject, and asked how they’d categorize it post-surgery—if it’s a raccoon or a skunk.

Their answers are rated on a scale of 1 to 3. In this system, if children shift their categorization—if they assert that the shaved raccoon is now a skunk, as it resembles this animal in appearance and smell—they’re given a score of 1.

In contrast, if children are unsure of how the transformed thing should be categorized—if it’s some “ra-kunk” hybrid of the two—they’re given a score of 2.

Finally, if children don’t change the initial categorization of the item—saying that the animal remains a raccoon, despite its altered physical state—they’re assigned a score of 3.

Here, the dependent variables are the average scores children receive across natural kind and artifact stories, which can be used to gauge children’s tendency to resist changing their identification of transformed objects.

Based on research performed by Frank Keil and colleagues, it is expected that children will readily change their categorization of inanimate artifacts, but not of living natural kinds, presumably due to some internal, unchangeable feature that animals possess.

Before the experiment begins, create cards with images of pre- and post-transformation items. Design 16 such transformation vignettes—eight dealing with artifacts, and eight dealing with natural kinds.

When the child arrives, greet them and introduce them to the concepts of doctors and surgeries.

Present the 16 transformation vignettes to them in a random order: “The doctors took a coffeepot that looked like this. They sawed off the handle, sealed the top, took off the top knob, sealed the spout, and sawed it off. They also sawed off the base and attached a flat piece of metal. They attached a little stick, cut a window in it, and filled the metal container with bird food. When they were done, it looked like this.”

“After the operation, was this a coffeepot or a birdfeeder?”

If in any instance the child gives an ambiguous or hybrid response—like saying that a coffeepot transformed into a birdfeeder is now a “coffee-feeder”—also ask them to call it by either its pre- or post-transformation name. This encourages children to think about their answers, as well as warrants different responses across all of the vignettes.

Question the child in a free-form manner about the reasons behind their response, in order to determine the features or principles of the post-transformation item that led them to their conclusion.

For each vignette, record the child’s response for future analysis.

When the child has completed all 16 vignettes, have two independent raters read their responses and code them on a scale of 1–3.

To analyze the data for each of the two types of transformations—either natural kinds or artifacts—plot the average children’s score as a function of age.

Perform an analysis of variance to determine if there are differences between the three age groups or two types of transformations.

Notice that as age increases, the average scores of children for the artifact transformations remained relatively constant, hovering around 1.25. This indicates that children at all ages studied readily change their categorization of altered artifacts.

In addition, these results suggest that the perceptual features of post-transformation items in these instances—what a child sees—informs their categorical placement. In contrast, for natural kind transformations—especially those involving animals—children’s average scores increased as a function of age. This indicates that as children get older, they gradually represent category membership as an internal, unchangeable aspect of animals.

Now that you know how researchers are using transformation scenarios to better understand how children categorize general artifacts and natural kinds, let’s look at other applications of this technique.

As you’ve learned, natural kinds can be either living items—like plants or animals—or inanimate objects, like minerals.

Thus, some researchers are further looking into how children come to categorize these different types of natural kinds, and if any differences exist between these processes.

Such work has demonstrated that whereas children quickly learn to categorize animate natural kinds—and resist the tendency to change their categorization of animals post-transformation—a similar ability for inanimate natural kinds develops more slowly.

Some psychologists have suggested that parental influence may play a role in this phenomenon. For example, a parent may emphasize that an animal should be categorized as a fish if its parents were fish, and if it lays eggs that also yield fish.

However, given the advanced subject matter of what makes an inanimate natural kind—like the organization of electrons in its atoms, or the elements that constitute it—parents may not have similar discussions with their children about such items.

Other researchers are looking at how children apply facts about a single member of a category to make inferences about other items in the same category.

For example, psychologists could tell children that a mouse is more active at night and likes to eat cheese, and then show them a collection of realistic pictures composed of mice with different coats, other rodents, and unrelated animals—like a cow.

By either including or removing labels from these images, researchers can evaluate the extent to which textual category labels, in addition to morphological features, factor in children’s ability to infer that other mice have the same characteristics.

You’ve just watched JoVE’s video on how children develop the ability to categorize—and make inferences about—natural kinds and artifacts. By now, you should know how to design and use transformation stories to investigate this phenomenon, and collect and interpret results. You should also have an understanding of how psychologists are using this method to investigate other aspects of the categorization process.

Thanks for watching!