James O. Coplien, Software Production Research
Maria Villar, PhD
Organizations improve their performance by learning from experience and by assessing the changing context in which they execute. Most organizational improvement programs are based on process specifications. A process can be effective as a learning tool only if it captures the values of the organization using it as part of a systemic whole. We propose a research program to explore the value systems suitable to an effective development process. The research explores several questions, including whether a process can be made general enough to be portable across several value systems. The proposal builds on standard techniques from sociology, anthropology, and other fields of human study. The work also builds on previous industry research using social, psychological and anthropological techniques.Keywords: process, values, patterns, culture, reflection, principles
Contemporary process description rarely capture all components of learning. Learning has three major components: learning from cause and effect; identifying principles that link cause and effect; and reconciling practice with values of the development community [Swieringa+1992]. Current process descriptions usually capture cause-and-effect relationships; they rarely capture principles, and almost never capture values. We believe that by providing process models at the principle and value level, we can provide appropriate and powerful tools to the process designer, provide better alignment criteria while assessing how well a given process fits a given organization (process acceptance), and smooth deployment as processes evolve and extend to new organizations. A value- or principle-based process is less constraining than one that encodes the explicit context of cause-effect pairs, so it can capture the living structure of an organization as it rises to the challenges of changes in the market, workplace, work force, etc. In the extreme, it may be possible to derive processes anew from (first) principles and values rather than evolving each organization's process from a generic starting point. Such an approach can leverage local culture to improve employee loyalty and a sense of personal value-added, which in turn can fuel major gains in productivity and quality.
This research will lead to a new methodology for formulating, deploying and evolving processes. We anticipate that the new method will go beyond process-writing in two dimensions: the front end and back end. The front end will incorporate more grounding in principles and values. The back-end will deploy processes in a form different from what we use today, perhaps using casuistic patterns (see "values" in the glossary) and articulated value systems extracted from the client organization itself.
More generally, this proposal explores broad and fundamental issues of contemporary thinking about software development process. The link from values, to organizational structure, and finally to development processes will give management new tools to support both their employees and their customers.
How do mission/vision statements shape everyday (practiced) values?
These studies found deeper problems. If one takes the standard industry definition of process,
a set of partially ordered steps intended to reach a goal [Curtis1992]then most development jobs (like load building) could in fact not be described in process form. Most processes don't exhibit the regularity and repeatability necessary for formalization; those that do can usually be automated, as was done in the PRL project to automate database integrity-checking rule generation [ICCL1994]. This is a critical weakness of the process approach, since all process improvement depends fundamentally on compliance between the model and actual practice, and on repeatability of actual practice.
These studies concluded that conventional process is not a good basis to understand or improve processes. Disciplines that study cultural behavior don't think in terms of "steps intended to reach a goal," but view goals as a value "with a specifically operational effect" and view "values [as] govern[ing] behavior" ([Hohmann1996], p. 190; see "goal" in the glossary). Hall ([Hall1962], p. 283) notes:
The ultimate objective is to understand and predict human behavior... Since systems engineers are (supposed to be) men of action, we need to know if psychological values can help us solve problems. In particular, do they help in choosing good objectives? The answer is obviously, yes.Cockburn's value maps: Alistair Cockburn (Humans and Technology) presented a panel position at OOPSLA '96 describing the role of value maps in establishing organizational direction and process [Fraser+1996]. The work hasn't yet been formalized, but the interest in this work in a forum like OOPSLA underscores the community concern and interest with organizational issues that go beyond process to organization and values.
Psychology of Organizational Change: Change and organizational learning are duals of each other. To understand how an organization learns, it's important to study cultural evolution: the institution of new values, the merging of cultures and their value systems, and occasional conflicts that arise from value system clashes between cultures that interface with each other.
There has been much good work in the psychology of change. Weinberg [Weinberg1996] builds on the change models of Virginia Satir, who originally developed them for family therapy. The Satir model makes explicit links between phases of organizational learning during change and individuals' psychological makeup (e.g., values).
Complexity Theory: Another emerging change model is that of emergent organizational behavior, relevant to organizations in constant churn. Contemporary research and practice in post-modern sociology has demonstrated the success of techniques that have analogues in chaos theory. Leading industry advocates of this work include McMasters [McMaster1995] and Wheatley [Wheatley1994]. McMasters' thesis is that complex systems such as human systems often exhibit emergent behavior generated by a small number of simple principles. These principles link closely to values in our research, and they correspond closely to the patterns of the organizational pattern community (see immediately below). Wheatley goes beyond McMasters' models and takes the "strange attractors" of chaos theory as direct metaphors for the few simple principles that drive complex behavior.
Organizational pattern work: Process and organization research in Bell Labs Research over the past five years has corroborated the weak correspondence between empirical practice and the written process specification ([Cain+1993]). The work has identified numerous patterns (see glossary) that bode well for success in software development cultures [Harrison+1996]. Empirically, the research found consistently distinguished social structures in those organizations that stood out for supporting values such as product quality and productivity. This work helped de-couple software development success from one of its commonly supposed drivers--design paradigm--and tied it to deeper patterns (see glossary) and principles.
Not only do these patterns seem to be suitable for judging important values (those related to teamwork, effectiveness, productivity, and employee satisfaction) in existing organizations, but they can guide dysfunctional organizations in paths that can better satisfy the organization's values. Experience has borne out the success of this approach in practice [Coplien1994]. In this sense, patterns are a proven alternative to process formulation and introduction.
These patterns have strong parallels in the pattern taxonomies of classic anthropology (e.g., [Kroeber1963], Chapter 8). Kroeber speaks of universal patterns that transcend cultures; of systemic patterns that vary in detail from culture to culture, but which have common roots in some single common historical source; and of total culture patterns that give a culture its identity. One finds all three of these kinds of patterns in [Coplien1994].
Swieringa and Wierdsma [Swieringa+1992] have elaborated a 3-tier model for organizational learning. The model recurs in other literature, e.g. in Sengé [Sengé1990] and others, and is consistent with the statistical process control approach to organizational learning of Deming. The three basic levels are:
The Learning Discipline: The work of Sengé [Sengé1990] has wide-ranging implications for the role of patterns and value systems in organizational development. He notes that process models that depend on a linear view of time, and the relationship between the events along that line, leads to a shallow model of single-loop learning:
What, exactly, does it mean to say that structures generate particular patterns of behavior? ([Sengé1990], p. 45)This suggests (our opinion) that we should look for signs of polychronic orientation in the development cultures we study, and how it clashes with the monochronic outlook of process standards. This may correspond closely to a clash between polychronic orientation of development organizations and the monotonic orientation of management culture.. . .a fundamental characteristic of complex human systems . . . [is that] "cause" and "effect" are not close in time and space. By "effects," I mean the obvious symptoms that indicate that there are problems--drug abuse, unemployment, starving children, falling orders, and sagging profits. By "cause" I mean the interaction of the underlying system that is most responsible for generating the symptoms, and which, if recognized, could lead to changes producing lasting improvement. Why is this a problem? Because most of us assume they are--most of us assume, most of the time, that cause and effect are close in time and space. ([Sengé1990], p. 63)
Systems Engineering: Hall's ([Hall1962]) foundational work on systems engineering in the Bell System, came at a time when the technical community gave the value system prominent stature. He described long-term planning at the system level, based on an assessment of enterprise values. These values were driven by the theories both of economic value (business economics) and psychological value (psychometrics), and to a lesser degree by the theory of casuistic value (cultural continuity). Targeted to a technical community, Hall's work is grounded in social, anthropological, and psychological theory.
Hall doesn't so much offer empirical findings as he reflects the techniques of an era of technological maturity where explicit consideration of value was still part of the Bell System culture. His work contains pointers to many tools (e.g., for establishing value order scales and interval scales) that can be useful in our research.
Allen's work on social distance: Thomas Allen is another noteworthy researcher whose work crosses over from social analyses to first-hand concerns of the technical community. at Sloan School has done voluminous studies on the structure of technical organizations, particularly as it relates to the physical architecture and geographical situation of the population's work places [Allen1977]. If we choose to investigate the role of geographical distribution on culture, and what that portends for process portability and uniformity, Allen's research lays important groundwork.
The interesting question here is: does structure come from values as well? Do different cultures have different values that lead to different norms of social distance?
Alexander's moral imperative: Christopher Alexander's work in architecture [Alexander+1977] has had a strong influence on the software development community, particularly on that part of it concerned with object-oriented programming. In particular, his work has set a tone that makes these communities more conscious of issues of morality and values. This has important impact on two levels. It helps practitioners in our fields see themselves as people who are making the world better through their work, which gives them a sense of purpose and contribution. But even before that, the community itself needs to struggle with its sense of purpose and its identity and, in the process, pursue the higher levels of learning and development suggested by Swieringa and Wierdsma [Swieringa+1992] in their model of organizations learning.
From Alexander's recent keynote address at OOPSLA '96:
And what I am proposing here is something a little bit different from that which is a view of programming as the natural genetic infrastructure of a living world which you are capable of creating, managing, making available. And which could then have the result that a living structure in our towns, houses, work places, cities, is an attainable thing which it has not been for the last 50 to 100 years. That is an incredible vision of the future. I realize that you probably think I'm nuts! Because this is not what I'm supposed to be talking about to you. And you may say, well, gosh, great idea, but we're not interested. But I do think you are capable of that. And I don't think anybody else is going to do this job [create a sense of care for wholeness and aesthetics in the world through our work].
In general, only an interdisciplinary Research program can rise above the asymptotic approaches to success that characterize contemporary efforts. Research focus is traditionally narrow. Since the important problems are systemic problems, a Research program must take a broad scope if its effects are to be broad, lasting and positive. In particular, this Research exercise will leverage insights from anthropology, sociology, and psychology, as well as from computer science.
The research program has three phases. The first phase is a pre-study phase where the team will gain understanding of the tools each other has found effective in their research. This phase is particularly important because of the diverse backgrounds of the team members. The second phase is the experiment itself, which includes developing a research plan, collecting data through interviews, etc. We combine the design and data-collection stages to allow for strategic changes in the research methodology. Last is the analysis and documentation phase, where we extract structure from our findings, report our findings to our stakeholders, and publish results as appropriate.
Anthropology attempts to understand the human condition in the broadest sense possible. It examines all aspects of human nature and behavior, explores similarities and differences across space and time, and seeks to produce generalizations about humankind. This research programs shares the major goals of Anthropology, as much as its empirical concerns. We want to conduct holistic analyses, examine contextual forces, and explore differences in values and understandings among individuals, and within and between groups. We propose three workshops to discuss conceptual issues and potential research obstacles and to examine available methodological tools.
| Premise | Tool or Approach |
|---|---|
| What appears to be a single cultural system (e.g., a project) may be multiple cultures, each driven by different values (e.g. management, coder, testers, and marketers) | Pasteur social network analysis [Harrison+1996] and psychometric studies using game theory or other approaches to establish an order scale of value, using tools such as that of Von Neumann and Morgenstern ([Hall1962], ff. 286) |
| Mismatch between the norms of psychological value within the culture, and those articulated by the corresponding management culture | Psychometric studies to establish an order scale of value. |
| Sengé's hypothesis ([Sengé1990]) that problems and solutions are not close in time and space, while most processes assume that they are. | Anthropological assessment of monochronic or polychronic orientation in individual groups and in the population as a whole. MBTI personality assessments may also prove useful, looking for SJ preferences in the process/management culture, and NP preferences in the work culture. |
| That geography is a dominant factor in shaping social interactions, more so than any institutional organization structure | Xenophobic tendencies within social groups and geographically centralized social units, and a measure of social distance between these social units (social network analysis using the Pasteur method as in [Harrison+1996]) |
Using these and possible other tools, we will investigate interacting value systems in software organizations. Our main target organization right now is ASOS, though we may engage other organizations as resources permit and as interest arises from such organizations. We will explore interacting value systems in each project, and will analyze individual and group values to test premises and, hopefully, identify patterns that characterize or transcend organizations.
Organizational culture is the pattern of basic assumptions that a given group has invented, discovered, or developed in learning to cope with its problems of external adaptation and internal integration, and that have worked well enough to be considered valid, and, therefore, to be taught to new members as the correct way to perceive, think, and feel in relation to these problems. (Schein)Kroeber ([Kroeber1963]) notes that it's more important to understand what culture is not, rather than what it is.
...culture might be defined as all the activities and non-physiological products of human personalities that are not automatically reflex or instinct. That in turn means...that culture consists of conditioned or learned activities (plus the manufactured results of these); and the idea of learning brings us back again to what is socially transmitted, what is received from tradition, what "is acquired by man as a member of societies." So perhaps how it comes to be is really more distinctive of culture than what it is.Hohmann ([Hohmann1996]) defines culture as comprising:
| Value: Take care of employee. | Goal: Remain profitable, so a fair wage can be paid and job security will be enhanced. |
| Value: Staying current with new technology is important. | Goal: Provide at least two weeks of training per year in an aspect of technology considered "new." |
| Value: Provide superior customer service. | Goal: Control growth within the company so each client is given appropriate attention. |
pattern: In the parlance of classic anthropology, patterns are "those arrangements or systems of internal relationship which give to any culture its coherence or plan, and keep it from being a mere accumulation of random bits" ([Kroeber1963], p. 118). Some patterns, called universal patterns, are common to all humanity. Systemic patterns are fundamental utilitarian structures within a culture that can vary from culture to culture, but which can all be traced back to a single root. Total culture patterns are those that identify and characterize a culture (German culture, Italian culture, etc.)
In recent years, the software development community has turned to patterns as a way of capturing knowledge about repeatedly successful structures and practices. These patterns were inspired by pattern work in the field of building architecture, published in the well-known works of Christopher Alexander (e.g., [Alexander+1977]). Some of these software patterns describe cultural solutions in software development organizations (e.g., [Coplien1995]). These patterns differ from the patterns of classic cultural anthropology in that they are goal-driven, in that they presume causes (called forces and context) which elicit them, and that they drive toward a well-defined, desirable outcome. That is, these patterns presume a value system. The software pattern culture pays explicit homage to its value system.
psychometrics: the discipline of measuring preferences, values, impressions, feelings, etc., particularly as it relates to the psychological theory of value. Psychometrics help researchers establish a value scale (see "value", below) for value system control variables.
value: Hohmann ([Hohmann1996], p. 190) notes that "a value governs behavior and is expressed through culture." Values are the focus of this research.
A philosophical system of values has absolutes. It may place material matters lowest and spiritual values highest. This is one of the common senses of value, but not the one we're studying here. Another perspective is the social perspective, which categorizes values as economic, moral, truth, political, ethical, aesthetic, and religious. We will be examining values from a psychological perspective, which looks at human needs and tendencies.
Hall distinguishes between the economic theory of value, the psychological theory of value, and the casuistic theory of value, three noteworthy examples of a range of values. He writes ([Hall1962], p.252):
While there is no general theory of value, there are, as noted in connection with the classification of decision criteria...a number of special theories of value... By comparing these theories of value we hope, at least, to advance along the road to answering these questions: (1) What is the nature of values? (2) What are the values in systems engineering, and how do these relate to the fundamental values of life in general? (3) How can these values be determined and used to make good decisions, especially good choices of objectives?Economic value is based on what one must give up to get something else; it always involves a choice and trade-off. It includes value-in-exchange (market value) and value-in-use. Economic market value is a zero-sum game, based on a shared community concept of value. Value-in-use denotes the importance an individual places on one thing relative to another in relation to personal wants or needs.
Psychological value is defined in terms of feeling. In general, value resides in "any sort of interest or appreciation" ([Hall1962], p. 282) where appreciation, or other feelings, are the gauge of value. For psychological value, value and the feeling of value are the same thing. Psychological value can be assessed with a number of tools from psychometric research.
Casuistic value is based on past experience, and is best typified by the legal profession and by organizations that build on their folklore. It's a bit different than the contemporary pattern discipline which dissects past practices to apply them analytically (through analysis of context and forces) instead of applying the practices per se.
[Allen1977] Allen, Thomas. Managing the Flow of Technology. Boston: MIT Press,
1977, 141-182.
[Cain+1993] Cain, Brendan G., and James O. Coplien. A Role-Based Empirical Process
Modeling Environment. Proceedings of Second International Conference on the
Software Process (ICSP-2), Los Alamitos, California: IEEE Computer Press, February, 1993.
[Coplien1994] Coplien, J. The Human Side of Patterns. C++ Report 8(1), SIGS Publications, January 1996, 81-85.
[Coplien1995] Coplien, J. A Development Process Generative Pattern Language. In Coplien, J. O., and D. Schmidt, eds. Pattern Languages of Program Design. Reading,
MA.: Addison-Wesley, 1995.
[Curtis1992] Curtis, Bill, Marc I. Kellner and Jim Over. "Process Modeling." Communications of the ACM 35, 9, September 1992, 75-90.
[Fraser+1996] Fraser, Steven, Alistair Cockburn, Leo Brajkovich, Jim Coplien, Larry
Constantine, and Dave West. OO Anthropology: Crossing the Chasm. ACM SIGPLAN Notices 31(1), October 1996, 286-291.
[Hall1962] Hall, Arthur D. A Methodology for Systems Engineering. Princeton, NJ:
Van Nostrand, ©1962.
[Harrison+1996] Harrison, N. B, and J. O. Coplien. Patterns of Productive Software Organizations. Bell Labs Technical Journal 1(1), Summer 1996, 138-145.
[Hohmann1996] Hohmann, Luke. The Journey of the Software Professional. Upper
Saddle River, NJ: Prentice-Hall, ©1996.
[ICCL1994] Ladd, D. A., and J. C. Ramming. A*: A Language for Implementing Language Processors. In Proceedings of the International Conference on Computer
Languages, May 1994.
[Kroeber1963] Kroeber, A. L. Anthropology: Culture, Patterns and Process. New York:
Harcourt, Brace and World, ©1948, Printed 1963.
[McMaster1995] McMaster, Michael D. The Intelligence Advantage: Organizing for
Complexity, ©1994-1995, Knowledge Based Development Company, Ltd., Celtic
House, Victoria Street, Douglas, Isle of Man.
[Sengé1990] Sengé, Peter. The Fifth Discipline: The Art and Practice of the Learning
Organization. New York: Doubleday, ©1990.
[Swieringa+1992] Swieringa, Joop, and André Wierdsma. Becoming a Learning Organization. Reading MA: Addison-Wesley, ©1992.