Category Archives: Social psychology

Teaching Bias, Part 1

Prejudice and discrimination, Social psychology,

You may be surprised to hear that White children show evidence of bias against African-Americans as early as age 3. How does this happen? Since there is evidence that the implicit biases of adults leak out through their nonverbal behavior, it seems reasonable that children pick up these cues from their parents and older acquaintances. A new study by Allison Skinner and her colleagues shows how exposure to positive or negative nonverbal cues can create social biases in preschool children. The studies are simple, but they are awkward to explain, so please bear with me.

In the first experiment, 67 4- and 5-year-old pre-school children watched a video in which two adult female actors each exhibited nonverbal bias toward two adult female targets. The targets were idenified by the colors of their shirts, red or black. Although the actors used exactly the same scripts when talking to the two targets, one target received positive nonverbal signals (i.e., smiling, warm tone of voice, leaning in) and the other received negative signals (scowling, cold tone, leaning away). Since these were two different women, the actual identity of the targets who received the warmer and colder treatment was counterbalanced; that is, each woman received positive and negative treatment an equal number of times over the course of the experiment.

After the video, the researchers gave the children four tasks designed to measure which target they preferred. The first was a simple preference question asking which woman they liked better. For the second, they were given an opportunity to behave prosocially. They were asked to which target the experimenter should give a toy. The two remaining tasks were opportunities to imitate one of the two targets. In the third task, they had to choose which of two labels to give to a toy which the two targets had called by different names. In the fourth, they had to choose one of two actions, ways to use a cone-shaped object, which the two targets had used differently.

The children showed a preference for the target who received the positive nonverbal treatment on three of the four tasks—all but the action imitation. A summary measure of the number of times out of four they showed favoritism toward the preferred target was statistically significant.

The researchers were less interested in demonstrating favoritism toward specific individuals than in the development of favoristism toward groups of people. In a second experiment, they measured whether the preferences demonstrated in Study 1 would generalize to other members of the target’s group. The two targets were introduced as members of the red group and the black group, matching the colors of their shirts. After the video, the children were given three tasks—preference, prosocial behavior and label imitation. The results replicated those of the first study.

Then the children were introduced to two new adult woman targets, said to be members of the red and black groups (wearing appropriate-colored shirts), who were best friends of the previous two women. They were asked which friend they liked better and were asked to imitate the actions of one of the two friends. The results showed greater liking for and more imitation of the friends of the preferred target on the video. In other words, the favoritism toward the preferred target (and against the non-preferred target) generalized to other members of their groups.

This study is a demonstration experiment. To prove that they understand how a widget works, researchers will show that they can create a widget in the laboratory. The widget in this case is group favoritism. We should not be put off by the fact that the groups are artificial, defined only by the colors of their shirts. Suppose the researchers had used members of real groups, such as White and African-American women, as their targets. In that case, the researchers would not have created group biases, since 4- and 5-year-olds already have racial attitudes. For evidence of how pre-existing attitudes can be strengthened or weakened by the way targets are treated, please read Part 2 of this post.

You may also be interested in reading:

What Does a Welfare Recipient Look Like?

Racial Profiling in Preschool

A Darker Side of Politics

So Far, It Looks Like It Was the Racism

Politics, Prejudice and discrimination, Social psychology, , , ,

One question has dominated the conversation among political scientists attempting to explain the presidential election: Were Trump’s supporters motivated primarily by economic anxiety or racial resentment? So far, I’ve avoided weighing in on this question, hoping that the definitive study would appear. It hasn’t yet, but a new experiment by Michael Tesler is interesting enough to warrant giving you a progress report.

The corporate media narrative clearly favors the economic explanation. In a typical article, we are told (correctly) that the family incomes of working class families have been stagnant for 35 years, that trade agreements and the 2008 recession have caused widespread unemployment and underemployment, and that both political parties have ignored the plight of these Americans. This is followed by interviews with a couple of Trump supporters who express pain and anger over the way they have been treated. However, this is anecdotal evidence. The answers given by Trump supporters are partially driven by the questions they are asked. For the media, framing the election in terms of economic anxiety rather than racism avoids offending Trump and his supporters.

Much of the evidence available prior to the election failed to support the economic anxiety narrative. Surveys showed that racial attitudes predicted Trump support better than economic attitudes—for example, these two, and this one. This large sample Gallup poll also cast doubt on the economic explanation. The median household income of a Trump supporter in the primaries was $72,000, higher than the median income of Clinton supporters ($61,000) and the general population ($56,000). In addition, post-election analyses showed that Clinton received more votes in economically-distressed communities—those with a higher percentage of their population below the poverty line.

Michael Tesler has been studying the racialization of politics for over a decade. Racialization refers to the tendency of racial attitudes to influence opinions toward a variety of other issues not obviously related to race, such as health care or gay marriage. Tesler embedded an experiment within a YouGov/Huffington Post national survey of 1000 voters conducted on December 6 and 7. Half the participants were asked if they agreed with the following statement:

Over the past few years, Blacks have gotten less than they deserve.

Ths is an item from the Symbolic Racism Scale, which is used to measure racial resentment. The remaining respondents were presented with this statement.

Over the past few years, average Americans have gotten less than they deserve.

Most people assume an “average American” is White. In 2005, Devos and Banaji conducted a series of five studies showing that the category “American” is more strongly associated with the category “White” than either “African-American” or “Asian-American.” Based on this evidence, Tesler assumed that respondents would interpret the second statement as referring to Whites. He then compared the responses of people who reported that they had voted for Clinton and Trump to these two questions.

This study pits the economic anxiety and racial resentment explanations against one another. Would Trump voters be more likely than Clinton voters to agree that average Americans have gotten less than they deserve? Or would differences emerge only when the question referred to Black Americans?

The results on the left show a typical racial divide between Black and White respondents. White participants were more than twice as likely to think that average Americans had gotten less than they deserve than to think that Blacks had gotten less than they deserve. Black participants thought everyone had gotten less than they deserve. Since there were more White than Black participants, the averages for the full sample resembled those of Whites.

The data on the right address the research question. Clinton voters were almost as likely (57%) to say that average Americans have gotten less than they deserve as Trump voters (64%). Since this was a large sample, this 7% difference is probably statistically significant, but it is small in comparison to the difference on the racial resentment item. Only 12% of Trump supporters agreed that Blacks had gotten less than they deserved, compared to 57% of Clinton supporters—a difference of 45%. The data are more consistent with the racial resentment interpretation of Trump’s victory.

Tesler frames the responses of Trump supporters as an example of the ultimate attribution error. Attribution is the processes by which we infer the causes of behavior. The ultimate attribution error is the tendency to take personal credit for our own successful behavior and that of our in-group, and blaming our failures on environmental obstacles, while at the same time blaming members of out-groups for their failures, and attributing their successes to unfair advantages. Given this bias, it follows that Whites have gotten less than they deserve, while Blacks have gotten more.

Were the election results caused by economic anxiety or racism?  We still await a more definitive study. It will require a larger sample of voters and a valid measure of economic anxiety, with statistical controls for other variables known to influence voting decisions. If I see such a study, I’ll let you know.

You may also be interested in reading:

Trump’s Trump Card

What Does a Welfare Recipient Look Like?

Framing the Debates

The Hidden Injuries of (the) Class(room)

Education, Social psychology,

This is the first of what I expect to be a series of posts on The Hidden Injuries of Class, named for the 1973 book of that title by Jonathan Cobb and Richard Sennett.

It is obvious that education is one of the primary institutions that reproduces social inequality; that is, it is one of the main reasons there is so little social mobility in the United States. Most commentators attribute this to two reasons.

The current research focuses on a third possibility—class differences that emerge from within the classroom. Since UMC children are more familiar with classroom norms, more “education-ready,” the classroom is not a level playing field, but rather one in which UMC children perform better while WC children appear more withdrawn and less intelligent. One classroom tradition that elicits these class differences in behavior asking students to raise their hands. Hand-raising introduces competition and social comparison into the classroom environment, which can be detrimental to the self-esteem of WC kids.

Sebastien Goudeau and Jean-Claude Croizet conducted three experiments with fith and sixth grade French school children. Children were divided into those with WC and UMC backgrounds according to the status of the parent with the highest occupational level. In the first study, 953 students responded to a reading comprehension test in which 15 questions were read and posted on a screen and the children wrote their answers is a notebook. The number of correct answers was scored. In a randomly determined half of the classes, students were instructed to raise their hands when they thought they knew the answer. In the other classes, this instruction was omitted.

The results showed that hand-raising interfered with the performance of the WC children but had no effect on the UMC kids.

In the other two experiments, the children’s actual social class was not a variable. Instead, the authors attempted to create a cultural advantage for half the children. The students performed a coding task in which they were asked to write down symbols that had been associated with certain letters. Half the students were given fifteen practice trials to become familiar with the coding task, while the others were given only five practice trials. Each class contained a mixture of students who were more or less familiar with the coding task.

As before, in half the classes, the kids were asked to raise their hands when they knew the answer, and in the other half, they were not. The results showed that hand-raising disrupted the performance of the students who were less familiar with the task, but had no effect on those more familiar with it.

Part of the reason social class differences are so discouraging to WC children is their hidden nature, which creates the illusion that there are real differences in ability between WC and UMC children. In the third study, all classes were instructed to raise their hands when they knew the answer, but an additional variable was introduced. Half of the classes were correctly informed that some students had had more opportunity to practice than others, while the other classes were not. When the students who were less familiar with the task were informed that others had this advantage, hand-raising no longer disrupted their performance.

Godeau and Croizet propose that UMC children enter school with more cultural capital than WC children. Parents provide their children with social capital when they teach them the skills and attitudes they need to succeed in school. Certain classroom practices, such as asking children to read aloud in front of the class, will serve to increase these differences. Asking teachers to avoid these activities is unlikely to be successful, and might deny WC kids an opportunity to overcome their deficits. More promising might be a preschool experience in which WC children are familiarized with educational culture. They should  be told often that their unfamiliarity with these behaviors does not mean that they are stupid.

You may also be interested in reading:

Racial Profiling in Preschool

Asian-American Achievement as a Self-Fulfilling Prophecy

Did Ebola Influence the 2014 Elections?

Media, Politics, Social psychology, , ,

Republicans did very well on Election Day 2014, gaining control of the Senate for the first time in eight years and increasing their majority in the House of Representatives. Most pundits attributed these results to low turnout by Democrats in a non-presidential election year and to President Obama’s poor approval ratings, due primarily to the disastrous rollout of the Affordable Care Act earlier that year. But a recent paper by Alec Beall and two other psychologists at the University of British Columbia suggests that breaking news about the Ebola epidemic also played a significant role in the election outcome.

Their paper contains two studies, both of which are interrupted time series designs. In this design, data that are routinely collected are examined to see if they change after a specific event. In the first study, they analyzed the aggregate results of all polls conducted between September 1 and November 1, 2014 that asked respondents whether they intended to vote for a Democrat or a Republican in their upcoming House election. The “interruption” occurred when Center for Disease Control announced the first Ebola case in the U. S. on September 30. The research question was whether the poll results changed from before to after that date.

The above results show support for the Republican candidate minus support for the Democratic candidate in the month and the week before and after the Ebola story broke. In both cases, the temporal trends were significantly different from before to after September 30. The before and after lines had different slopes, and the shift was in favor of the Republican candidates. The authors also collected data from Google on the daily search volume for the term “Ebola,” and found that it too was positively related to Republican voting intentions.

Beall and his colleagues examined two possible alternative explanations—concern about terrorism and the economy. They measured daily search volume for the term “ISIS,” and checked the Dow-Jones Industrial Average, which was dropping at the time. Interest in ISIS was (surprisingly) negatively related to Republican voting intentions and the stock market had no significant effect.

In their second study, the authors looked at the 34 Senate races. They computed 34 state-specific polling averages by subtracting Democratic voting intentions from Republican intentions. Then they subtracted the September results from the October results. Thus, a higher number would indicate a shift toward the Republican candidate. The aggregate results showed a significant increase in Republican voting intentions after September 30.

However, not all states shifted in the same direction. Using Cook’s Partisan Voter Index, they determined whether each state had voted more for Republicans or Democrats in recent years. Then they analyzed the data separately for “red” and “blue” states. The results are shown below.

The changes were in the direction of the state’s dominant political party. In the red states, the Republican candidate did better after September 30. In the blue states, the Ebola scare seemed to help the Democrat, although the effect was smaller. This could also be interpreted as a shift toward the favorite, since candidates who were leading before September 30 tended to do even better after that date.

This study is part of a small but increasing body of research which shows that external threats that cause fear in the population seem to work to the advantage of conservative political candidates. In a previous post, I reported on a British study which indicated that the 2005 London bombings increased prejudice toward Muslims. More to the point is a 2004 study in which reminding participants of the 9/11 terrorist attack on the World Trade Center increased support for President George W. Bush in his campaign against John Kerry. These studies are consistent with older research suggesting that social threats are associated with an increase in authoritarianism in the U. S. population. Authoritarian attitudes are characterized by obedience to authority, hostility toward minority groups and a high degree of conformity to social norms.

Surprisingly, Beall and his colleagues did not mention terror management theory as a way of understanding their results. According to this theory, human awareness of the inevitability of death—called mortality salience—creates existential terror and the need to manage this terror. One way people manage terror is through defensive efforts to validate their own cultural world views—those attitudes that give their lives meaning and purpose. Previous research suggests that mortality salience results primarily in conservative shifts in attitudes, including support for harsher punishment for moral transgressors, increased attachment to charismatic leaders, and increases in religiosity and patriotism. (A charismatic leader is one whose influence depends on citizen identification with the leader or the nation-state, as in “Make America great again.”) The Bush v. Kerry study mentioned in the preceding paragraph was intended to be a test of terror management theory.

One of the effects of saturation coverage of the Ebola epidemic was to remind people of the possibility of their own death and that of loved ones. The results of the 2014 House elections are consistent with a terror management interpretation. The Senate results do not contradict the theory, since there was an overall shift in favor of Republican candidates, but they add an additional detail. In states that usually voted Democratic, the Ebola scare increased support for Democrats. If mortality salience causes people to reaffirm their cultural world views, this could have produced a shift toward liberalism in states in which the majority of citizens held progressive attitudes.

Research findings such as these suggest the possibility that political parties and the corporate media might strategically exaggerate threats in order to influence the outcomes of elections. Willer found that government-issued terror alerts between 2001 and 2004 were associated with stronger approval ratings of President Bush. Tom Ridge, Director of Homeland Security at the time, later admitted that he was pressured by the White House to increase the threat level before the 2004 election. Since that time, it has become routine for Republicans to emphasize threats to the public’s well-being more than Democrats, and evidence from the 2016 presidential debates suggests that the media gave greater attention to Republican issues.

Republicans made Ebola an issue in the 2014 election, claiming that President Obama was failing to adequately protect public health and arguing that he should close the borders and not allow Americans suffering from the virus back into the country for treatment. In retrospect, news coverage of the threat of Ebola appears to have created unnecessary panic. Analysis of the motives of the media decision makers is complicated by the knowledge that they also exaggerate threats because they believe that increasing public fear leads to higher ratings. Media Matters for America presented data showing that coverage of Ebola plummeted immediately after the 2014 election was over (see below). However, I know of no “smoking gun” showing that the corporate media deliberately created panic in order to help Republican candidates.

You may also be interested in reading:

Are Terrorists Getting What They Want?

Framing the Debates

Trump’s Trump Card

The Stroking Community, Part 2

Education, Social psychology, ,

Please read The Stroking Community, Part 1 before continuing.

The Grading Leniency Assumption

The evidence for the bias assumption questions the validity of SETs, but it does not, by itself, explain grade inflation. The grading leniency assumption adds that college teachers try to obtain favorable evaluations by assigning higher grades and by reducing course workloads. Stroebe cites three surveys that show that a majority of faculty believe that higher grades and lower workloads result in higher SETs. One survey published in 1980 found that 38% of faculty respondents admitted lowering the difficulty level of their courses as a result of SETs. (I’m not aware of any more recent survey which asked this question, which is unfortunate.)

It should be noted, of course, that faculty may not be aware of having changed their behavior, or they may think they have done it for other reasons. One common reason given for watering down courses is that contemporary students are unprepared for college-level work. (One former colleague, for example, said, “You have to meet students at a place where they feel comfortable.” Unfortunately, that “place” gets closer to the downtown bars with each passing year.)

Indirect evidence for the grading leniency assumption comes from student behavior. Greenwald and Gillmore note that students would ordinarily be expected to report working harder in courses in which they expect to get a higher grade. However, in a study of over 500 classes, students reported doing less work in those courses in which they expected to get a higher grade, a finding which is readily explained by the grading leniency assumption.

Finally, there are studies of the effects of grades on future course enrollment. Some universities publish average grades by course and instructor at the university’s website, and it is possible to determine through computer signatures whether students have accessed this information. In two studies, consulting past grading data predicted future choices of courses and sections, with the sections with higher grades being preferred by about 2 to 1. In one of these studies, this preference for easier courses was greater among low ability students than high ability students.

It should be noted that lowering the students’ workload not only improves faculty evaluations, it also lowers the faculty’s own workload. There are fewer of those time-consuming term papers and essay exams to grade. Instead, teachers can give the multiple-choice exams that are considerately provided free of charge by the textbook publisher.

The faculty members with the most to lose in the current enviroment are those who attempt to maintain high academic standards and are punished for their integrity with low student evaluations. If they don’t have tenure, they could be fired. And even if they do have tenure, they are likely to be under considerable pressure from administrators to improve their evaluations.

Grade Inflation

Here’s another chart to remind you of how bad grade inflation has gotten. It shows the change over time in the frequency of letter grades.

Grade inflation is an unintended consequence of universities’ reliance on student evaluations. Can it be considered a good thing? Kohn proposes that grades serve three functions: sorting, motivation and feedback. If grades gradually lose their meaning, they become less useful as sorting criteria for employers and graduate schools and less useful as feedback to students. The students most harmed are the hard-working, high ability students who would have gotten A’s in the absence of grade inflation. They are no longer able to distinguish themselves from their more mediocre colleagues. Leading average students to believe they are doing better than they actually are could lead to unpleasant shocks after they graduate.

The motivational function of grading assumes that the rewards and punishments provided by grades induce students to work harder and learn more. But the picture that emerges from the course selection studies is one of students attempting to obtain higher grades without working for them. Stroebe suggests that grade inflation is most likely to demotivate high ability students, who might decide that studying is not worth the effort if they wind up with the same grades as their less deserving classmates.

It’s hard to see how grade inflation can be reversed. The Wellesley solution of mandating lower grades holds some promise, but only if it is adopted by almost all similar universities at about the same time, since if some universities attempt to control grade inflation while others do not, their students will be at a competitive disadvantage when applying for jobs or to graduate school. Princeton initiated a similar program, but abandoned it after peer colleges failed to follow suit. There was some concern that controlling grade inflation might cause studients not to come to Princeton.

A shorter-term solution is suggested by Greenwald and Gillmore. They propose that SETs be statistically corrected for the average grade in the class. Although their method is complicated, the gist of it is that if the distribution of grades in a class is lenient, SETs are reduced. If the distribution is strict, the instructor receives a bonus. Although this makes good sense to me, it’s hard to imagine a university faculty agreeing to it.

The implications of this research are depressing. Students and professors are rewarding one another for working less hard. They are caught in a social trap in which short-term positive reinforcement serves to maintain behavior that has long-term negative consequences for themselves, the university and the society. Meanwhile, colleges and universities, already under financial stress, are decaying from the inside out because they are failing to meet their most basic obligation—that of helping and requiring students to learn.

You may also be interested in reading:

The Stroking Community, Part 1

Asian-American Achievement as a Self-Fulfilling Prophecy

Racial Profiling in Preschool

The Stroking Community, Part 1

Education, Social psychology, ,

Grade inflation has been a fact of life at American universities for several decades. College grades are measured on a 4-point scale (A = 4, B = 3, C = 2, D = 1, F = 0). Since the 1980’s, grades at a large sample of colleges and universities, have increased on average by .10 to .15 points per decade. The overall grade point average now stands at about 3.15.

This would seem to imply that students have either gotten smarter or are working harder. However, verbal SAT scores of incoming students have declined sharply during this period, while math scores have remained relatively stable. There has also been a decline in the amount of time students report that they spend studying. On average, college students now claim to study only 12 to 14 hours per week. Assuming 16 hours of class time, that amounts to a work week of less than 30 hours.

More disturbing is the research of Richard Arum and Josipa Roksa. They administered the Collegiate Learning Assessment, a cognitive test measuring critical thinking, complex reasoning and writing, to 2300 students at 24 universities in their first semester and at the end of their sophomore year. They found only limited improvement (.18 of a standard deviation, on average), and no improvement at all among 45% of the students. Of the behaviors they measured, only time spent studying was associated with cognitive gains.

Beginning in the 1980s, colleges and universities entered what is sometimes called the “student-as-consumer” era. Almost all of them began routinely administering student evaluations of teaching (SETs), and basing decisions about tenure and promotion of faculty members in part on their SETs. Social psychologist Wolfgang Stroebe, in an article entitled “Why Good Teaching Evaluations May Reward Bad Teaching,” argues that SETs are responsible for some of the grade inflation. Stroebe has organized the research on SETs around two hypotheses which he calls the bias assumption and the grading leniency assumption.

The Bias Assumption

It has long been known that higher student grades are associated with better evaluations, both within and between classes. That is, within a class, the students with the highest grades give the instructor the most favorable evaluations. When you compare different classes, those with the highest average grades also have the highest average SETs. A recent meta-analysis found that grades account for about 10% of the variability in teaching evaluations.

Since these data are correlational, their meaning is ambiguous. They were initially interpreted to mean that teaching effectiveness influences both grades and evaluations. If so, SETs are a valid measure of instructional quality. Stroebe’s bias assumption states that students give favorable evaluations in appreciation for having less work to do and higher grades, and that this is an important source of bias which undermines the validity of SETs.

Over the years, this debate has been a source of animosity among college faculty. It is probably the case that SET believers receive more favorable evaluations than SET skeptics. SET believers sometimes accuse SET skeptics of making excuses for their poor student evaluations, while skeptics suggest that believers are in denial about the possibility that their high ratings are obtained in part by ingratiating their students.

The obvious—but unethical—way to test the bias hypothesis is to manipulate students’ grades in order to see what effect this has on SETs. Back in the days before ethical review of research with human subjects became routine, there were a few studies that temporarily gave students false feedback about their grades. They found that grades did affect evaluations. For example, in one study, students in two large sections of General Psychology taught by the same instructor were graded on slightly different scales. The instructor received better evaluations in the section with the more generous grading scale.

There are several other research findings that, while correlational, are consistent with the bias hypothesis.

  • In the early 2000s, Wellesley College, concerned about grade inflation, instituted a policy requiring that average grades in introductory course be no higher than 3.33. This resulted in an immediate decline in grades. Average SETs declined significantly in the affected courses and departments.
  • Greenwald and Gillmore found that the grade a student expected affected not only ratings of teaching effectiveness, but also had significant effects on logically irrelevant factors such as ratings of the instructor’s handwriting, the audibility of his or her voice, and the quality of the classroom. This suggests that there is a general halo effect surrounding lenient instructors.
  • The website Rate My Professors (RMP) contains 15 million ratings of 1.4 miliion professors at 7000 colleges. Professors are rated on easiness, helpfulness, clarity, “hotness” and overall quality. Easiness—a question that is seldom asked on institutional evaluations—is defined on the website as the ability to get a high grade without working hard. RMP ratings closely match the institutional SET scores of the same professors. The two dimensions most highly correlated with overall quality are easiness (r = .62) and hotness (r = .64). Obviously, the professor’s physical attractiveness is another threat to the validity of student evaluations that is deserving of study.

In my judgment, the best test of whether teachers with high evaluations are really better teachers are those studies which examine the effects of SETs in one course on performance in a followup course. For example, do students who give their Calculus I instructor a high rating do better in a Calculus II course taught by a different instructor? Stroebe found five studies using this research design. Three of them reported that those students who gave their instructors high SETs in the first course did more poorly in the followup course, one of them found no difference, and the fifth reported a mixture of negative and null effects depending on the item.

Merely finding no relationship between SETs in Course 1 and grades in Course 2 raises questions about the validity of SETs. The negative relationship found in the majority of these studies has a more radical implication. It implies that students learn less from those teachers to whom they give high evaluations. One of these studies found, however, that ratings of grading strictness in the first course were positively related to performance in the second.

Its important to noted that Stroebe does not claim that SETs are totally invalid as measures of teaching effectiveness, but only that they are strongly biased. Poor student evaluations can serve as a warning that faculty are not meeting their obligations. One recent study found a non-linear relationship between SETs and its measure of student learning. The students learned the most from professors whose SETs were near the middle of the distribution. They learned the least from those whose evaluations were the lowest and the highest.

There are a number of possible explanations for the bias hypothesis. One is simple reciprocity. When a professor does something nice for a student, the student returns the favor with a positive evaluation. SETs give students who are unhappy with their grades an opportunity to exact their revenge. A second explanation for the negative ratings given by students with lower grades is attributional bias. The self-serving attribution bias predicts that we maintain our self-esteem by taking personal credit for our successful behaviors but blaming our failures on external causes, such as poor teaching or unfair grading by the professor.

Please continue reading Part 2.

You may also be interested in reading:

The Stroking Community, Part 2

Asian-American Achievement as a Self-Fulfilling Prophecy

Racial Profiling in Preschool

What Does a Welfare Recipient Look Like?

Politics, Prejudice and discrimination, Social psychology,

Economic inequality in the United States is at record levels. In surveys, Americans say they would prefer a more equal distribution of wealth. However, the majority consistently votes against public assistance programs that redistribute wealth. Political scientist Martin Gilens, in his 1999 book Why Americans Hate Welfare, attributes this primarily to racial prejudice. Gilens examined the photographs that accompanied stories about poverty in the news magazines Time, Newsweek and U. S. News. African-Americans accounted for 62% of the poor people shown in the photos. On the ABC, CBS and NBC nightly news programs, 65% of poor people shown in reports on poverty were Black. In reality, as of 2010, 32% of welfare recipients were Black, 32% were White and 30% were Hispanic.

Gilens also did an experiment in which a “welfare mother” was identified as either White or Black. Participants who read about a Black welfare recipient were more opposed to welfare than those reading of a White recipient. The implication of Gilens’ research is that White Americans’ disdain for welfare is explained in part by racial prejudice. Americans hate welfare because they overestimate the percentage of recipients who are African-Americans. However, there is a missing link in this analysis. Gilens implies, but does not show, that Americans are influenced by these misleading media reports—that is, that the average American’s mental image of a welfare recipient is a Black person.

A research team headed by Jazmin Brown-Iannuzzi of the University of Kentucky sought to measure their participants’ mental representations of a typical welfare recipient using an unusual technique. I’m not sure I completely understand it without seeing a demonstration, but the image generation phase of their study goes something like this: First, they constructed a computer-generated “base face,” a composite of a Black man, a Black Woman, a White man and a White woman. Then, on each of 400 trials, the computer introduced noise which altered the base image in two opposite directions. The participants were asked to choose which of these two altered faces most resembled a welfare recipient and which one resembled a non-welfare recipient. (Race was never mentioned.) The computer then generated a composite image of a typical welfare recipient and a typical non-welfare recipient, based on all the responses of all the participants.

This was done twice, with 118 college students participating in Study 1 and 238 internet volunteers in Study 2. The composite faces from the two studies are similar and are shown below. Although the composite faces of the welfare recipients look like African-Americans, I presume this was less apparent to the participants as they made their 400 decisions.

During the second phase of these two studies, 90 different participants were shown one of the composite faces and were asked to rate the person on a number of different dimensions. No mention of welfare was made to these participants. The raters judged the welfare recipient composites as more likely to be African-American (rather than White) than the non-welfare recipient composites. The welfare recipients were also rated more negatively on 11 different traits, including lazier, more incompetent, more hostile, less likeable and less human(!). These studies fill in the missing link in Gilens’ research. The average person’s mental image of a typical welfare recipient is of an African-American.

Finally, Brown-Iannuzzi and her colleagues did a third study, an experiment in which 229 internet volunteers were shown one of the composite images—either a welfare recipient or a non-welfare recipient—and asked a number of questions. The critical items were whether they would support giving this person food stamps and cash assistance. The other questions repeated some of the ratings used in the previous studies. Here are the results. This study replicates the Gilens experiment mentioned in the second paragraph.

In summary, the first two studies showed that when asked to imagine a typical welfare recipient, people generate a mental image of an African-American, while their mental image of a non-welfare recipient is that of a White person. The third study demonstrated that when other people are shown these mental images, they were less supportive of giving welfare to the composite typical welfare recipients than the composite non-welfare recipients.

Finally, the authors did a mediational analysis to see which variables mediated between the composite images and the decision to support or not support giving welfare to that person. The data were consistent with the following causal chain (see below): The image leads first to an inference that the person is either Black or White. This, in turn, leads to a judgment of how deserving the person is. (Black people are less deserving.) Finally, the judgment of deservingness leads the decision of whether to support giving welfare to the person.

We are going through a period of extreme racialization of politics. Americans’ racial attitudes influence their opinions about other political issues that may or may not be related to race. In some cases, survey participants’ racial attitudes determine their attitude toward a policy merely because they believe President Obama does or does not support the policy. Not only do racial attitudes appear to have been the strongest predictor of support for Donald Trump, they mattered more in electing Trump than Obama.

Nowhere is racialization more evident than in attitudes toward financial relief for the poor. People support income redistribution in principle, but they overestimate the percentage of poor people who are Black. As a result, their racial prejudice discourages them from supporting income redistribution policies.

You may also be interested in reading:

Old-Fashioned Racism

The Singer, Not the Song

Racialization and “Student-Athletes”

“Here I Am. Do You See Me?”

Cognitive psychology, Social psychology,

Maybe because of the continuing increase in economic inequality in the United States, social psychologists are taking a greater interest in social class differences in behavior. I have previously written about studies showing that upper class people are less likely to help a person in need than lower class individuals, and are more likely to engage in unethical behavior—behavior that is potentially harmful to others.

A new article by Pia Dietze and Eric Knowles of New York University suggests an explanation for these differences: Upper class individuals regard others as less motivationally relevant—that is, less “potentially rewarding, threatening or otherwise worth attending to”—than lower class members do. If that is the case, then members of the upper clases should pay less attention to other people they meet in public places.

In the first of three studies, Dietze and Knowles asked 61 college students to take a walk around the streets of Manhattan “testing” the Google Glass, a device that fits over the right eye and records what the person is looking at. Six independent judges watched these videos and measured the participants’ social gazes—the number and duration of their looks at the people they passed. The students were asked to classify themselves as either poor, working class, middle class, upper-middle class or upper class. These five labels were treated as a 5-point continuous scale.

Results showed that the number of social gazes did not differ by social class, perhaps indicating that it is necessary to at least glance at passers-by to successfully navigate the sidewalk. However, as predicted, the higher the self-reported social class of the participants, the longer the time they spent looking at the people they passed.

Is this only because other people are less “motivationally relevant” to upper class participants? After reading this study, I thought about sociologist Erving Goffman’s concept of civil inattention. Goffman said that when we pass strangers, we glance at them briefly, but then quickly look away, in order to avoid the appearance of staring at them. Are upper class children more likely to have been taught that it’s not polite to stare? Fortunately, the other two studies the authors report don’t involve face-to-face interaction and are not subject to this alternative explanation.

In the second study, 158 participants were asked to look at several visually diverse street scenes while fitted with an eye-tracking device which measured which part of the scene they were looking at, and for how long. The authors recorded the time spent looking at both people and things (cars, buildings, etc.) in the environment. Time spent looking at things did not differ significantly by social class, but participants who classified themselves in the lower classes spent more time looking at people. This is illustrated in the chart below, which compares working class and upper-middle class participants. (Study 2a involved 41 New York City scenes, while Study 2b added an equal number of scenes from London and San Francisco.)

In the last study, 397 paid internet volunteers participated in a flicker task. On each trial, participants were shown two rapidly alternating slides consisting of pictures of a person’s face and five other objects. On some trials, the two slides were identical, but on others, one of the six pictures—either the person or one of the five things—was different. Participants were asked to press a key as quickly as possible indicating whether the slides were the same or different, and the computer measured how rapidly participants responded. It was expected that lower class participants would be better at detecting changes among the people, but not among the things. This hypothesis was confirmed.

Although the flicker task has no obvious relevance to everyday life, the fact that the lower class participants detected changes in the faces more rapidly than the upper class participants suggests that they were more likely to be looking at the faces, rather than some other part of the slide. The fact that the differences were in milliseconds—a millisecond is a thousandth of a second—suggests that this is an automatic response rather than one that is under conscious control.

The chart above is from an article by Michael Kraus and two colleagues summarizing research on class differences in behavior. The present studies deal with cognition. Lower class people’s cognition is said to be contextual because it is directed at the social environment, probably because their lives are controlled more by outside forces, such as bosses and government policies. Upper class people are more likely to be paying attention to themselves and their own thoughts. It is hypothesized that this explains the differences in prosocial (helpful) behavior among the lower classes vs. selfish behavior among the upper classes that I noted in the opening paragraph. It may also help to explain class differences in political party affiliation and voting behavior, as long as voters are not confused or misled about which policies the candidates actually favor.

You may also be interested in reading:

Class Act

Me First

Racial Profiling in Preschool

Name That Cognitive Bias

Cognitive psychology, Social psychology

Cognitive and social psychologists have identified an amazing number of cognitive biases to which we are all prone, especially when we are not thinking slowly. I realize the chart is too small to read. You can look at a larger version here.

cog-bias

This reminds me of an old joke.

Patient:  “Doctor, I’m suffering from a nameless dread.”

Therapist:  “Not to worry. We have a name for everything.”

Racial Profiling in Preschool

Education, Prejudice and discrimination, Social psychology, ,

Data from the U. S. Department of Justice, Office of Civil Rights, shows that African-American children, especially boys, are suspended or expelled from preschools at a higher rate than White children. For example, while 20% of preschool boys are Black, 45% of the boys suspended are Black. However, this is not proof of racial discrimination, since a skeptic could argue that, even at this young age, Black children are more likely to misbehave.

A new study by Walter Gilliam and his colleagues at the Yale University Child Study Center takes an experimental approach to this issue by holding the behavior of Black and White children constant and observing how teachers respond. Participants were 132 prechool teachers recruited at an annual conference. Sixty-seven percent of the teachers were White and 22% were Black. They took part in two studies.

In the first study, participants were shown a 6-minute video of four preschool children—a Black boy, a Black girl, a White boy and a White girl—seated around a table. The teachers were asked to watch for “challenging behavior,” but in fact the video did not contain any misbehavior. A computerized eye-tracking device was used to measure the amount of time the teachers spent watching each child. At the conclusion, the teachers were asked to report which of the four children required the most attention.

The eye tracking results showed that the participants spent more time looking at boys than girls, and more time looking at Black children than White children. In addition, the time spent gazing at the Black boy was significantly greater than would have been expected on the basis of his combined race and gender. The race of the teacher made no difference in this study.

The title of the paper frames the research as a study of implicit bias, and media reports of the study have followed suit. The authors define implicit bias as the “automatic and unconscious stereotypes that drive people to behave and make decisions in certain ways.” However, the teachers’ conscious appraisal of which child they paid the most attention to appeared to match the eye-tracking results fairly closely, as shown in the chart below. Apparently the teachers were well aware that they were paying more attention to the Black boy.

yale_implicit_bias_infographic_v07

I mention this because the term “implicit bias” is sometimes used to deny personal responsibility for one’s own and others’ discriminatory behavior on the grounds that it is unconscious. By labeling this as a study of implicit bias, the authors may have given their teacher-participants less blame for their behavior than they actually deserved.

In my title, I described these results as similar to racial profiling. Racial profiling targets people based on stereotypes about their race, as when the police stop and frisk Black teenagers having no evidence that they are committing crimes. Like the police, these teachers were scanning for misbehavior, and they responded by giving special attention to African-American boys. (An editorial writer for the New York Times drew this same analogy.)

These same participants also took part in a second experiment. In this study, they were asked to read a vignette describing a preschool child who repeatedly engaged in disruptive behavior. The child’s race and gender were manipulated by changing the child’s name (DeShawn, Jake, Latoya or Emily). Half the participants in each race and gender condition also read background information suggesting that the child lived with a single mother who was under a great deal of stress. The others were not given background information. The teachers were then asked to rate the severity of the child’s behavior and to recommend whether the child should be suspended or expelled.

The following results were found for ratings of the severity of the behavior.

  • The same behavior was rated as more seriously disruptive when the child was White than when he or she was Black.
  • Giving teachers background information increased the ratings of the severity of the behavior.
  • The Black teachers rated the behavior as more serious than the White teachers.
  • The background information increased the perceived severity of the behavior when the teacher was of a different race than the child, but the teachers responded more sympathetically to it when the teacher and the child were of the same race.

With regard to suspension or expulsion, the only finding was that Black teachers were more likely to recommend these options.

The results of the second study are not a good fit with the Department of Justice data, since the teachers appear to be discriminating against the White children. The researchers explain this by suggesting that these teachers expected the Black children to be disruptive, but held the White children to a higher standard. Therefore, the identical behavior was seen as more serious when attributed to a White child.

My guess is that had the same behavior been rated more disruptive when when the child was Black, the results would have been interpreted in a straightforward manner as discrimination against African-Americans. However, since the results were unexpected, a more complex explanation was presented. This explanation may be correct, of course. There is some evidence for “shifting standards” with respect to race. However, the authors could have strengthened their argument with a followup study measuring teachers’ expectations about the misbehavior of Black and White children and the extent to which the behavior described in their vignette violated those expectations.

Since the Black teachers were stricter overall, it appears that increasing the representation of Black teachers will not by itself reduce the number of suspensions and expulsions.

Some additional perspective on this issue is provided by a set of two experiments by Jason Okonofua and Jennifer Eberhardt. Their participants, grade school teachers, read a desription of either a White or Black boy in middle school who committed two infractions—one class disruption and one act of insubordination. After each incident, they were asked how severely the child should be disciplined.

There was no difference in the punishment recommended for Black and White boys after the first infraction. As shown in the table, the recommended disciplinary action increased in severity after the second infraction, but it did so more for the Black boy than for the White boy. (In the table, “feeling troubled” refers to a combined measure the the severity of the misbehavior and the extent to which it would hinder and irritate a teacher.)

Apparently, the teachers were more likely to infer a disposition to misbehave from two bad actions when the child was African-American than when he was White.

You may also be interested in reading:

White Prejudice Affects Black Death Rates

Outrage

Asian-American Achievement as a Self-Fulfilling Prophecy