Showing papers in "Systems Research and Behavioral Science in 1969"

Differential effects on response bias of computer vs. conventional administration of a social science questionnaire: An exploratory methodological experiment

Abstract: An experiment was performed to determine the bias-reducing effect of administering a social science questionnaire by computer. It was assumed-that subjects would regard typing answers directly into a computer as a situation which guaranteed them a greater sense of privacy and anonymity than the conventional situation wherein questionnaires or psychological tests are filled out by hand and then scrutinized, scored, and interpreted directly by other human beings. Based upon this assumption, two specific hypotheses were formulated: first, whenever the content of a question is regarded by a subject as highly personal and possibly disturbing, he will respond with greater honesty and candor under computer administration as opposed to conventional questionnaire administration; second, whenever an impersonal and emotionally neutral question is asked, no such difference in response tendency will occur. Results of the experiment tended to support both hypotheses.

Critique and comment: Proof of a theorem on majority rule

Abstract: A mathematical proof is presented of a result due to Douglas Rae: simple majority rule is the unique decision rule (from among all possible decision rules, k = 1, 2, -, n in a committee of size n) which minimized the probability that a typical member will support a motion which the committee rejects or oppose a motion which is accepted, if he assumes that future preferences are unknown and that members vote (for or against) independently.

On the structure of shared awareness

Abstract: Sentences of the form, “A thinks B thinks ,” arise in the analysis of game-like situations and in the sociology of knowledge. These are modeled by a Boolean algebra with operators, where “A believes x,” is written Ax. Operators ∩j=1 (∞i, Ai)j representing complete consensus or “common opinion” are defined. Properties of these epistemic operators are derived from postulates. Sociological features of content, stability and change of shared awareness are discussed. Finally, the model is used to solve a class of 2-person 0-sum games in which the conventional requirements for mutual knowledge are relaxed.

Schizophrenia: Diagnosis and prognosis

Abstract: The Stanley R. Dean Research Award was established by the Fund for the Behavioral Sciences and is presented jointly with the American College of Psychiatrists to emphasize the importance of basic research toward an understanding of schizophrenia; each year, a scientist who has made an important contribution in this area is honored. Following is the text of a lecture delivered by Dr. Gabriel Langfeldt on the occasion of his receiving the seventh annual award of $2,000 at a seminar held in New Orleans, Louisiana, February 1, 1969, by the American College of Psychiatrists. Dr. Langfeldt, a pioneer in diagnostic criteria for schizophrenia, outlines the procedures which led to the recognition of schizophreniform, or schizophrenia-like, psychosis, an important step toward isolating the “true” schizophrenias.

A computerized approach to the processing and analysis of life histories obtained in sample surveys

Abstract: Life histories undeniably offer many advantages, especially in examining sequences of behavior. Their high cost and difficulties in analyzing large numbers have restricted the use of life histories to case studies. In a sample survey of 1640 men in Monterrey, Mexico, partial life histories were systematically obtained on residence, education, marital status and family formation, health impairment, and work. Problems of reliability of response, particularly recall error, are of special significance in the evaluation of life histories. Omission of changes in status, improper ordering and dating of changes were minimized by: systematic recording, in which all years had to be accounted for; facilitation of recall, primarily by moving back and forth among several important areas of respondent's life; checks on consistency of response, by means of similar questions in other parts of the interview schedule. In processing the data only changes in the content of each of the variables were coded and punched on cards. A special FORTRAN program was written to permit the reconstruction of the entire life histories on computer tape from the “change” data cards. The program also allows for the formation of new index variables and collapsing of others. Maximum flexibility of analysis is possible since any given recorded event in the life history can be “located” with reference to age and year, to other events and to the other information gathered in the questionnaire. In this way a closer approximation to causal analysis is made possible. Stochastic models and path analysis are examples of statistical techniques well suited for these kinds of data. The data gathering and processing procedures outlined in this article are applicable to many other kinds of investigation, ranging from political behavior to mental illness.

A study of subjective evaluation models

Abstract: The research reported explores some questions which arise when staff groups, such as operations research groups, are faced with obtaining a quantitative representation of an individual's or an organization's evaluation scheme. The study is concerned with the ability of professional hospital personnel to develop and use a subjective evaluation model in a reliable manner, the possible forms of such a model, and the extent to which the models used by the professionals reflected some recognizable organizational evaluation scheme.

Alcoholism, mental illness, and stress in 96 drivers causing fatal accidents

Abstract: Surviving and deceased drivers responsible for 96 fatal accidents during a 38-month period in Washtenaw County, Michigan, and a like number of matched controls were evaluated to determine the prevalence of alcoholism, psychopathology, and social stress, as well as other medical and social variables. Information was obtained through interviews of the drivers, their relatives, friends, and employers as well as from traffic and arrest records. Thirty-six of the 96 fatality drivers and three control drivers were chronic alcoholics. In addition, 11 fatality drivers and six control drivers were classified as ldquofrequent, high quantity usersrdquo of beverage alcohol as herein defined. The 96-driver fatality group differed significantly in many respects from the control group. However, the magnitude of discrepancy was often due to the presence of the 36 alcoholics in the fatality group.The 96 fatality drivers averaged significantly more previous accidents (1.2 versus 0.8) and moving violations (3.1 versus 1.7) than the control group, exhibited significantly more psychopathology (41 percent versus 17 percent) and social stress (52 percent versus 18 percent), and included more low-social-class persons (76 percent versus 54 percent) and fewer higher-social-class persons (5 percent versus 29 percent) than the control group.The 36 alcoholic fatality drivers were significantly overrepresented in fatal accidents involving only one vehicle and in those occurring at night. In addition, they had significantly more previous accidents, serious psychopathology (69 percent versus 25 percent) and social stress (72 percent versus 42 percent) than the 60 nonalcoholic fatality drivers.

Information effects on decision-making

Abstract: This paper summarizes the results of an experiment dealing with the effects of initial information and feedback on goal-setting and performance of the competing individuals. Three hypotheses were examined:. 1) More information will result in more accurate levels of goal-setting and decision-making. 2) The goals set at a given time are a function of previous goals, previous success in achieving the goals and learning experience about the environment. 3) The rate of learning of participants with less information will be higher than those with more information. The subjects were 48 graduate business students in six groups, eight individuals in each. There were four distinct types of initial information and feedback, so that every two persons of the eight in each group received one of the four information and feedback types. The results tend to support, in part, hypotheses 1 and 3. No support was found for hypothesis 2.

Behavior modification of an autistic child.

Abstract: An autistic child was observed while on a hospital ward. Observation revealed that the clinical staff gave considerably more attention to his bizarre, inappropriate behavior than to acceptable responses. Subsequently these two classes of behavior were observed and recorded by independent observers whose reliability averaged 85 percent. Next the child was given social and primary reinforcement for appropriate behaviors which included compliance to requests, comprehensible verbalizations, use of play materials, and the learning of academic skills. Self-hitting, junk verbalizations, and tantrums did not receive reinforcement. In order to demonstrate the role of contingent reinforcement, the contingencies were reversed for a four-day period. Following the initial reinforcement procedures, responses such as comprehensible verbalizations climbed from two percent to 46 percent while withdrawn responses and junk verbalizations disappeared. When the reinforcement contingencies were reversed, some appropriate responses dropped 50 percent and inappropriate responses increased. When the original contingencies were reinstituted, the behaviors returned to previous treatment levels.

Effect of uncertainty on risk taking in individual and group decisions

Abstract: Experimental gambling situations were used to test (1) the effects of uncertainty on individuals' willingness to take risk and (2) the effects of group discussion under conditions of positive, zero, and negative expected value. In all three experimental conditions individuals risked less money under uncertainty. The comparison of group with individual decisions showed results consistent with a model which proposes that group discussion enhances prior expected values and also results in a risk bias effect when uncertainty is present.

A geometric classification system for 2 × 2 interval-symmetric games

Abstract: Assessment of comparability among experimental game studies is aided by a two-dimensional geometric classification system for all 2 × 2 games which “look the same,” up to a linear transformation, to the two players. Each game is represented as a point on a plane whose two coordinates, r3 and r4, are readily calculated from the entries of the game matrix. Spatial location of the various games on the (r3, r4) plane is shown to be related to (a) the ordinal relationships among the payoffs, (b) the availability of alternation strategies, (c) the ordinal relationships among the joint outcomes, (d) the probability of a cooperative choice in the Prisoner's Dilemma, and (e) the set of perceived games which might arise from a given set of experimenter-defined payoffs for subjects attempting to maximize different linear combinations of own and other's payoffs. The system is easily adapted to classification of games on the basis of any interval-scale property of interest.

Primate frontal lesions and the temporal structure of behavior.

Abstract: Removal of the frontal cortex of primates results in a psychological deficit usually classified in terms of short-term memory. This classification is based on impairment in per formance of delayed response or alternation type tasks. The present experiments were un dertaken in order to utilize this frontal preparation to learn more about the short-term memory process as well as to delineate more precisely the frontal lobe deficit. Rhesus mon keys with lesions of the dorsolateral frontal cortex were tested in a situation where the classical alternation task with 5 second delay (Right-Left-Right-Left) was modified by in terposing a 15 second interval between each R-L couplet (R-L R-L ). This modification made it possible for monkeys with frontal lesions who had failed the classical task to perform with very few errors. Two additional alternation experiments, one interposing light and one interposing sound between trials, failed to produce such effective performance. These experiments suggest that proper temporal structuring, or parsing, of the stream of stimuli to which the organism is subject, is a more important variable in the mechanism of short-term memory than is the maintenance of a neural trace per se. They also suggest that the frontal cortex of primates is critically involved in the temporal structuring of behavior.

An interactive technique for the analysis of multivariate data

Abstract: IDEA, a computer program for Inductive Data Exploration and Analysis, is operational under SDC's O-32 time-sharing system. Using IDEA, a researcher, through on-line inter action with his data, is able to explore various ways of accounting for relationships that exist among his measurements. Instead of conventional poly-nominal regression coefficients, IDEA represents these relationships as decision trees whose form, shape, depth and so on are determined by the user from his data, using unique IDEA heuristics to guide him. Three case studies are reported in which IDEA is contrasted with discriminant analysis. A summary is made of a preliminary study using artificial data of the effect of measurement error on IDEA analysis.

Legislative districting by computer

Abstract: A system for legislative redistricting by computer simulation is described. Included is a discussion of the algorithmic logic used in the main districting program, computer implementation of the algorithm, and supporting programs useful throughout the entire redistricting process. This system was developed as a result of using computer redistricting algorithms in Iowa sub districting activities.

The determination and scaling of a bivariate utility function

Abstract: A considerable body of evidence exists which indicates a lack of a monotonic relationship between real rewards and the utility for the corresponding outcomes in an experimental game environment. This phenomenon we have called utility function confounding. It often appears that individuals are influenced in their choice behavior not only by their expected rewards, but also by the relative amount of their reward to that of the other party or parties to the game. The authors develop a bivariate utility function designed to accommodate both utility for own reward and for own reward relative to that of the other party. This model is then shown to be general enough to encompass several distinct types of utility function confounding which have appeared in experimental situations. While difficulties remain in the measures required of the model, the authors' main hope is that this neglected aspect of subjects' utilities will receive the attention that it deserves, to the benefit of more meaningful experimentation of interpersonal interactions.

Social power and communication in small groups

Abstract: “Social Power” was defined and manipulated, in the context of 5-person interactim, as the conditional probability that a communicative act would succeed once it had been initiated. Five subjects sent content-free signals to each other, which they were instructed to accumulate for an eventual payoff of 1 cent per signal received. The only way a subject could control his receipts was by varying his own sending behavior. Unknown to the subjects, the probability (π) that their signals would reach their intended target was controlled by the experimenter. Ten different patterns of π values in five person groups were studied: .3.5.5.5.7, .3.4.5.6.7, .3.3.5.7.7, .1.5.5.5.9, .1.3.5.7.9, .1.1.1.1.5, .1.5.5.5.5, .5.5.5.5.9, 5.9.9.9.9. These patterns provided variations in average group power (π) and variance of group power (σπ2). The results showed that a subject's receipts depended in a simple linear manner in his own π value relative to π. If an individual's share of the receipts is designated P and there are N individuals in the group then the equation P = (1/N) + b(N - 2/N - 1) (π - π) provided a reasonable fit to the receipt data. This result follows from the more molecular assumption that each individual asymptotically sends signals to another with a probability that depends linearly on the other's relative power position among the possible targets. If Pij is the probability that i sends to j and πi is the mean power of subjects other than i in the group, then Pij = (1/N - 1) + b(πj - πi). The degree of differentiation in receipts in a group thus depends on the slope b; here b = .169 provided the best fit to the individual sending proportions. The value (N - 2/N - 1)b = .126 for receipt data was compared with a slope of .284 found in a previous study with three person groups, supporting a conjecture that differentiation would decline with size of the group.

Computers in behavioral science: The decisional design and cyclic computation of span

Abstract: The SPAN method of voting or pooling is based on computer processing of quantitative decisions made by group members. The method may be studied through behavioral research and applied to political, economic, and other fields of decision making.

Interaction of subject and experimenter expectancy effects in a tone length discrimination task

Abstract: The purpose of this experiment was to measure and study the interaction of both subject and experimenter expectancy effects in a tone length discrimination task, and to uncover the extent to which facial or verbal cues were responsible for the transmission of these effects. The task consisted of comparing a standard tone to each of nine numbered tones in a comparison group presented randomly. In all conditions the standard tone matched the center tone of the comparison set in length. In Study 1, six groups of seven Ss and seven Es each were employed, for a total of 84 participants. Though covertly introducing expectancies to either S, E, or both, as to where in the progression the standard matched, both subject and experimenter expectancy effects were induced, and found to interact additively to sway the data in the direction of the expectancy when compared to naive controls. It was found that Ss in groups that did not confirm the E's expectancy were judged by the Es as not to have been cooperative in following directions. In Study 2, three groups of five Ss and five Es each were used for a total of 30 participants. In all groups S was naive and E was given a downward expectancy. It was found that the experimenter expectancy effect could be transmitted by either verbal or facial cues alone. Controls must be initiated to eliminate the cues responsible for expectancy effects from the psychological experiment.

Information effects in two mixed-motive games†

Abstract: The class of 2 × 2 games in which each player's payoffs are strictly ordered contains exactly 78 games. Of these 78, 12 are symmetric, that is, they “look alike” to both players. Four of these 12 games are dilemma-type games, two of which, Prisoner's Dilemma and Chicken, have been the subject of considerable experimental investigation. The remaining two games, Hero and Leader, have not been subject to experimental study. This paper reports the strategy choice behavior of pairs of naive subjects playing Hero and Leader under two information conditions, with and without knowledge of the other's payoffs. The results obtained in this study indicate that Hero is played more cooperatively than Leader where cooperative interaction is defined in terms of the frequency and distribution of Pareto outcomes between players in a pair. Also, both games are played more cooperatively when a player has knowledge of his own and the other's payoffs than when a player has knowledge only of his own payoffs.

Computers in behavioral science on a computer program that generates and queries kinship structures

Abstract: The present work is an exercise to generate complex, inter-related data structures from the simplest possible information source. The organization of the memory is such that updating and retrieval processes are easy to perform. To demonstrate the capabilities of the program, the first fifteen chapters of Robert Graves: The Greek Myths, Volume I were used as data source. One of the interesting results is that, in spite of the rather free mating pattern of the divine participants, Zeus and Aphrodite adhered to a Platonic relationship. Another is that Rhea is Cronus' sister, wife, grand-aunt and daughter-in-law's mother at the same time. The program, written in IPL-V in its present form, can accept a maximum of 500 separate individuals and will process an unlimited number of inquiries in one particular run. Non- perfect information is appropriately treated. Finally, possible extensions are briefly described.

Conflict Resolution within Economic Organizations

Abstract: Alternatives for reducing goal conflict within an organization are analyzed via a mathematical model. To bring the goals of managers into line with organizational goals, compensation via salary and profit sharing are examined as well as schemes for imposing profit constraints and using transfer pricing. Conclusions of the analysis can be summarized in four propositions, (a) A salary compensation plan will not motivate managers in a decentralized organization to maximize company profits. (b) Profit sharing can be used to increase both division profit and the satisfaction of the manager relative to a simple salary compensation plan, (c) However, a bonus compensation plan cannot yield maximum company profits, (d) The combination of transfer pricing and a profit sharing compensation plan can provide maximization of company profits under decentralization.

Critique and comment: Bases for psychiatric prognostic favorability judgments: Psycho physical power functions?

Abstract: A previously developed scale based on magnitude estimations made by 29 psychiatrists regarding judged prognostic favorability for 15 functional psychosis disorder classifications had been found to be surprisingly reliable and had been classified, on the basis of two criteria, as a prothetic continuum. The present investigation's purpose was to attempt to relate this scale in the psychophysical manner to validity indices derived from research literature. Power functions were found to describe, approximately, the relationships of this scale to three such validity measures: percentage improvement, length of stay in mental hospitals, and age upon admission to mental hospitals. The possibility that clinical and sensory judgment may be based upon similar psychological mechanisms was discussed.

Toward the study of biopolitics: a cross-sectional analysis of mortality rates.

Abstract: A new field of inquiry, which might be called biopolitics, would identify the many possible relationships between political and medical-biological variables in societal systems. One such proposition is that levels of foreign and domestic violence are a function of societal conditions and processes that become manifest in the form of deaths due to psychogenic causes, such as suicides and heart disease. To test this hypothesis, data for all significant causes of death—46 in all—are collected for 72 political units along with deathrates due to foreign and domestic violence. Eight measures of possible error are also included. The year 1960 is chosen, except for the conflict data, which span 1958 to 1960. Factor analysis, smallest space analysis, and cluster analysis are performed on the data. A Q-analysis among the 72 sampling units reveals that there are only about three strata. R-analysis of all 56 variables, in contrast, is heterogeneous, containing 14 factors with eigenvalues over the conventional 1.0 cutoff level. In the factor analysis, foreign and domestic conflict are independent of all other causes of death, including each other. Smallest space analysis and cluster analysis, however, provide results that approximate second-order factor analysis: foreign and domestic violence converge as part of an over-all medical underdevelopment cluster. Psychogenic causes of death, which emerge within the developed cluster, are thus unrelated to political casualties in the mid-twentieth century. War and civil strife are associated with lower health standards across 72 countries.

Social process modeling: A comparison of a live and computerized simulation

Abstract: The extent to which an entirely computerized simulation could recapture the essential decision-making processes involved in a complex, but controlled laboratory exercise using human subjects is examined. A series of analytic simulations of an exercise called PLANS were run on a digital computer. Beginning with simulations based upon entirely random decision processes, greater and greater degrees of “decision making” sophistication were gradually incorporated into the computerized versions of PLANS. Past efforts to simulate human psychological process on a computer—for example, chess playing and music writing programs—have been directed generally at improving the effectiveness of computer programs. In the present study, involving the simulation of social rather than purely psychological processes, the situation appears reversed: computer programs behaved too rationally and efficiently.

Threats and bargaining efficiency

Abstract: A major focus of gaming research has been the effect of the availability of threat on the joint efficiency of play in variable-sum bargaining games. This research employs an experimental instrument designed to avoid the pitfalls pointed out by critics of earlier research in the area. The resulting data support the proposition that the availability of threat reduces bargaining efficiency.

Computers in behavioral science: Computer scoring of sentence completion data

Abstract: The development of techniques for computer-based personality assessment from sentence completions is outlined. The One-Word Sentence Completion (OWSC) instrument was designed to elicit data suitable for machine processing, while retaining most of the advantages of a free-response format. Two operative scoring systems are described. The first employs a “dictionary” of 4366 weighted response words to yield 25 scores from a 90-item OWSC form. The second system utilized a complex word-root data reduction procedure and a bank of 892 generic roots to produce scores for 40 variables. Initial reliability data and normative sex differences are reported, and future development of the technique is discussed.

Alcohol, drinking patterns, and the psychological probability of success

Abstract: This study compared the effects of alcohol on estimates of success given by alcoholics and nonalcoholics. Theoretically, alcoholism has been associated with a learned ambivalence toward the consumption of alcohol. It follows that alcoholics are likely to judge that alcohol will have a conflicted (both favorable and unfavorable) effect on their estimates of success in the performance of skill and chance tasks. The alcoholic and nonalcoholic subjects were asked to estimate their degree of success under conditions in which they considered themselves very lucky, lucky, neither lucky nor unlucky but realistic, unlucky, and, very unlucky. Conflict after drinking was assumed to be present when subjects judged themselves relatively successful at the lucky end of the continuum (representing the favorable effects of alcohol) and relatively unsuccessful at the unlucky end of the continuum (representing the unfavorable effects of alcohol). Thirty nonpsychotic hospitalized alcoholics, thirty hospital employees who drank socially, and thirty hospitalized chronic schizophrenics who reported a history of moderate but not social drinking behavior, were employed in the study. Ten of the thirty subjects in each comparison group were randomly assigned to a nondrinking control group condition, were told that they would not be getting a drink of whiskey, and were then asked to give their psychological probability estimates. Ten more subjects from each of the three comparison groups were randomly assigned to a no-delay drinking condition in which they drank their drink of whiskey and immediately practiced the skill and chance tasks, and then gave their psychological probability estimates. The last ten subjects were assigned to a delay condition in which they drank their whiskey, performed some tasks not relevant to the present study, and then gave their psychological probability estimates. The three conditions respectively manipulated the psychological consequences of being denied a drink, the primarily psychological consequences of having just had a drink, and the interacting psychological and physiological consequences of having had a drink one-half hour before. The data were analyzed in a five-way analysis of variance independent on two dimensions and correlated on three. The results of this analysis, and of an orthogonal polynomial analysis of linear trend, tended to support the hypothesis. The alcoholics were most conflicted when they had just had a drink, and were least conflicted when they had been denied a drink. The normals were most conflicted at having been denied a drink and least conflicted following a drink consumed one-half hour before. The schizophrenics were most conflicted about the drink consumed one-half hour before giving their judgments and were least conflicted about the drink that they had just consumed. The responses of the schizophrenics, who were controlled but nonsocial drinkers, suggest limitations to social learning theories of drinking behavior. Parallels between psychological probability measures and projective tests were discussed. Trend analysis and interaction means were found to be useful in the evaluation of some higher-order interactions.