Communications of the ACM

Communications of the ACM

Multidisciplinary Research on the Datapath of the Computing Disciplines

By Antonio M. Lopez, Frederick G. Lopez, Robert W. Lent, Madonna G. Constantine
Communications of the ACM, December 2007, Vol. 50 No. 12, Pages 46-50
10.1145/1323688.1323693
Comments

Teaching computer organization and architecture to undergraduate students brings many basic computer concepts and relationships into focus. An example is how relatively minor changes to the organization and architecture of the processor's datapath can significantly improve performance, provided these changes are well thought out and scientifically tested [10]. On the other hand, changes that are not supported with scientific rigor can lead to catastrophic results, causing unintended performance problems.

Studying the datapath's organization and architecture also shows that performance levels are optimized when hardware and compiler software work together to prevent pipeline stalls. There are examples of program instruction sequences that, if given without compiler software support to the datapath hardware, even hardware with forwarding, will still cause a pipeline stall. However, the pipeline cannot stall when compiler software anticipates and accommodates hardware limitations by optimizing a program instruction sequence prior to furnishing it to the hardware.

Keeping this computer organization and architecture imagery in our mind's eye serves as an analogy for today's human pipeline problems in the computing disciplines. As in computer organization and architecture, when hardware alone is supposed to resolve sequencing problems in program instructions, yet still stall, the human pipeline can also stall when computing researchers (analogous to hardware) go it alone in trying to understand the datapath problems of the computing disciplines without first enlisting the expertise of behavioral/social science researchers (analogous to compiler software).

Here, we explore the need for multidisciplinary research—combining computing and behavioral/social science faculty—on the "datapath" of computing disciplines. Scientifically supported change is needed in light of the increasing number of stalls in the pipeline (such as the decline in the number of students enrolling in the computing disciplines and the increased dropout rates at each pipeline stage).

Undergraduate enrollment in computer science and computer engineering programs declined in 2005/2006 for the fourth straight year, according to a study released by the Computing Research Association (www.cra.org) [12]. Yet, according to the Bureau of Labor Statistics, U.S. Department of Labor [1], demand for people with a bachelor's degree in a computing discipline is increasing at a 27% or more annual growth rate through 2014. Embedded in the decline is the continued underrepresentation of women and minorities in the computing disciplines. The underrepresentation of women [2] and African-American students [9] in the computing disciplines was the focus of concern and study for many faculty members in the computing disciplines well before this general decline began in 2001/2002. However, due to the general decline in enrollment, such studies have taken on much more importance. Needed is better understanding of the background and psychosocial factors that attract students (including women and minorities) to (or repel them from) the computing disciplines.

The decreasing supply of students and the increasing demand for a skilled computer work force sparks action and debate in both academic and industry circles. Some faculty in the computing disciplines point to external factors (such as offshoring and the dot-com bust of 2000) as the cause for the decline of students majoring in the computing disciplines. Others point to internal factors more oriented to specific computing disciplines (such as datapath organization and architecture factors) as the cause. Unfortunately, some faculty members in the computing disciplines are investing in datapath intervention programs based on considerations (such as anecdotal evidence or small samples in localized observational studies) that lack a scientifically rigorous basis.

Some computer industry leaders have asked the U.S. Congress to increase the number of H1-B visas [5] to help increase the number of job applicants in the U.S. with degrees in the computing disciplines. Other organizations (such as the Institute of Electrical and Electronics Engineers, USA) suggest there are plenty of computing jobs in the U.S. and that the H1-B visa program actually hurts U.S. IT workers [4]. Adding to this debate is the fact that women and particular minority groups (such as African-Americans, Hispanics, and Native Americans) are still significantly underrepresented in the computing disciplines [3, 6].

Faculty members in the computing disciplines know their particular disciplines but are not trained as behavioral/social scientists. Many faculty members in the computing disciplines are not familiar with "people models," and their research interests rarely require them to submit proposals to their Institutional Review Board for Human Participants Study, (www.nsf.gov/bfa/dias/policy/human.jsp). However, when a human anomaly appears within the scope of their disciplines, some faculty members want to understand it and thus seek its cause. Some recognize their limitations as neophyte behavioral/social scientists, whereas others do not.

Observations of gender- and/or ethnically segregated career populations or subcultures raise a host of fascinating (sometimes controversial) questions and inquiry efforts. For example, Why do these differences exist? and Is it possible to facilitate more proportional participation of underrepresented groups in these activities? Such questions point up the idea that the identification of gender or ethnic differences on any dimension of human behavior explains only that a noteworthy difference in group representation/participation apparently exists. Research does not contribute substantially to an explanation of a human behavior unless and until it posits and measures theory-relevant variables that are presumed to account for the observation of "between group differences" in that behavior. It also must gather information on this variability from a range of data sources—human participants—in a variety of venues. When this research yields replicable patterns in the data that are consistent with posited underlying theoretical assumptions and constructs, researchers can begin to claim some understanding of the problem.

This point is relevant to the limitations of many studies on the topic of gender and ethnic underrepresentation in the computing disciplines conducted by faculty members solely from the computing disciplines and published in the computer-education literature. Many such studies have tried to shed light on the complex problem of underrepresentation by conducting non-theory-based, descriptive, cross-sectional research seeking to identify demographic/background variables (such as parental employment and number of math or programming courses taken) that may be associated with group differences. In addition, many such studies have relied on only limited samples drawn from single institutions and assessed at a single point in time. Such procedural decisions inevitably constrain the generalization of the findings. That is, because of these design and procedural constraints, research scientists cannot be confident that patterns observed in participants at a particular site are representative of stable, enduring response patterns in the larger population of students whose choice behavior they want to understand.

Back to Top

Datapath Correction

In the fall of 2003, the National Science Foundation awarded researchers at Xavier University of Louisiana a grant (HRD-0332780) to conduct a multidisciplinary study of gender-based differences and ethnic and cultural models in the computing disciplines. The grant allowed for the formation of a multidisciplinary research team of faculty in the computing disciplines and in psychology, primarily counseling. Counseling is one of the three major applied disciplines in professional psychology; clinical psychology and school psychology are the other two. Unlike clinical psychologists who favor the study and treatment of more psychologically impaired populations, counseling psychologists have historically favored the study of the psychological concerns and adaptations of otherwise normal-functioning populations.

As a consequence, counseling psychologists are interested in how people cope with normal life difficulties and transitions (such as those involving education and work decisions). They typically prefer theoretical and treatment models that are more holistic and contextual and that emphasize an understanding of person-environment interactions that contribute to problem functioning. Unlike school psychologists who work mainly in elementary and secondary education settings, counseling psychologists are generalists working in a range of professional settings, including college and university counseling centers.

Within professional psychology, counseling psychologists have made significant and sustained theoretical and research contributions to vocational/career psychology, as well as to the study of human diversity and cultural factors affecting human behavior. Indeed, the major theories of vocational choice/career development, as well as many of the major models of ethnic and cultural identity, are the work products of counseling psychologists. At the heart of the Xavier project was a particular contemporary counseling psychology product: social cognitive career theory (SCCT) [7], which focuses on the interplay among personal, environmental, and behavioral variables believed to influence a person's academic interests, choices, and performance outcomes. It seeks to explain the processes through which people develop basic academic and career interests, make and revise their educational and vocational plans, and perform in their chosen academic and career pursuits (see the figure here).

Self-efficacy (beliefs about one's ability to perform specific behaviors), coping-efficacy (beliefs about one's ability to deal with specific environmental barriers), outcome expectations (beliefs about the consequences of given actions), interests, and goals play key roles within each of the areas involving educational and career choices. They operate in concert with personal, contextual, and learning variables (such as gender, ethnicity, and social support and barriers) to help shape people's career trajectories. SCCT hypothesizes that students enter disciplines in which they express interest, believe they have the requisite abilities, and expect favorable outcomes [8].

In the Xavier project, we used measures of social cognitive person variables (such as self-efficacy and outcome expectations) and contextual variables (such as perceived social support and barriers) to predict students' interests and intentions to remain in computing (persistence goals). We hypothesized that over the length of the study—2004 to 2007—the set of SCCT variables would tend to predict students' interests and persistence goals across gender, ethnicity, and university type: historically black colleges and universities, HBCUs, or predominantly white institutions, PWIs. With only first-year data analyzed as of October 2007, preliminary findings from a path analysis indicate that the SCCT model fits the data very well, explaining roughly one-third of the variance in students' interests in computing tasks and intention to remain enrolled in one of the computing disciplines. However, this is a preliminary result and second- and third-year data must still be included in the analysis.

The Xavier project is also investigating a variety of additional psychological variables, some relating to SCCT and others to alternative theoretical perspectives. For instance, the project includes a focus on students' math and computer self-efficacy beliefs, collective self-esteem (the value we place on our own cultural group), perceived stereotype threat (the extent to which we are willing to ignore a socially held negative mental image of our cultural group), and gender roles. This interplay of concepts, relationships, and theory would not be possible if the various datapaths through the computing disciplines were viewed solely through the lenses of computing discipline faculty.

The interplay of a multidisciplinary team, including faculty in the computing disciplines and in the behavioral/social sciences, promises to yield a clearer understanding of the potential student pipelines in the computing disciplines, as well as of the reasons for the historic and continuing underrepresentation of women and minorities in some of the pipelines. Moreover, this multidisciplinary partnership is likely to help our counseling psychology colleagues develop a more nuanced appreciation of the computing disciplines. For example, although SCCT has been used to guide the study of student choice and persistence in science and engineering fields [8], findings may or may not generalize to the computing disciplines.

Findings in one computing discipline may or may not generalize to other computing disciplines. Computer engineering, computer science, software engineering, information systems, and information technology each have different goals, pedagogies, and domain-specific methodologies, leading to different bodies of knowledge, albeit with a common root in computing [11]. The learning experience of students entering one computing discipline differs significantly from students entering other computing disciplines.

Back to Top

Conclusion

Our longitudinal study (begun in the fall of 2004) was intended to introduce a new sample of first-year undergraduates in the fall of 2005 and the fall of 2006. When we wrote and submitted the grant proposal in 2002, the project goal was to survey students from 50 U.S. institutions of higher education, 25 HBCUs and 25 PWIs. For each year of the study, each institution was to recruit approximately 70 of their computing-discipline undergraduates (male and female); these research participants were to range from first-year students to graduating seniors. As an entering-undergraduate population baseline each year, 30 first-year undergraduates from noncomputing disciplines at each institution would also be surveyed. Faculty members at the 50 institutions would have the opportunity to form multidisciplinary teams consisting of one faculty member from the computing disciplines and one from a noncomputing discipline. The faculty members in the noncomputing disciplines included those in psychology, sociology, education, and women's studies. Some administrators, responsible for increasing diversity at their institutions were also expected to support the data-collection effort.

Little did any of us know in 2002 that the decline in enrollment in the computing disciplines, which was just beginning, would continue for at least the next five years. Furthermore, Hurricane Katrina, which would devastate New Orleans on August 29, 2005, forcing one of us to be away from Xavier University of Louisiana for five months, could not have been foreseen. Declining enrollment serves to underline the timeliness of this project. The hurricane considerably delayed the launch of the second year of the survey to the spring of 2006, at which point the number of first-year computing discipline students was low to nonexistent at many of the participating institutions.

However, 50 institutions of higher education were indeed involved in the second year of the study, including 23 HBCUs and 27 PWIs, where four PWIs had large Hispanic or Native American student populations. Analysis of the second-year data will add valuable evidence regarding the predictive utility of the variables and models under consideration. The third year of data collection took place in the spring of 2007. Additional information regarding findings from other multidisciplinary, longitudinal, large-sample, and multi-venue studies is in [3].

Back to Top

Back to Top

Back to Top

Back to top

The Digital Library is published by the Association for Computing Machinery. Copyright © 2007 ACM, Inc.

---

No entries found