Reference: Fuller, R.M. and Dennis, A.R. (2009), Does fit matter? The impact of task-technology fit and appropriation on team performance in repeated tasks, Information Systems Research 20(1), 2-17.
This is an important piece of research. It’s not often that new research dispels or substantially modifies well accepted models of how things work. This is one such example. Prior research has supported the intuitive belief that the fit between a technology and the task to which it is applied significantly affects success in performing the task. Presumably, the poorer the fit between a task and the technology used to perform it, the worse the performance of the task. This “Task-Technology Fit” theory was first formalized in MISQ in 1995 (Goodhue) and has been extensively analyzed, developed, and verified by subsequent research. However, as Fuller and Dennis show, the TTF theory is incorrect when applied to a repeated task. It turns out that by the third time the task is repeated, users will have figured out how to adapt the technology and the way the task is accomplished so that there is no significant difference in their performance or their perception of the technology. While poor-fit teams failed to equal the performance of well-fit teams in initial task performance, the differences between them melted away over time, becoming indistinguishable by the third repetition of the task!
This finding, while surprising and counter-intuitive, has some theoretical grounding. It is rooted in Adaptive Structuration Theory (AST) (Desanctis & Poole, 1994). AST holds that in performing a task, people adapt the elements of the tools that they use, the features they select, the rules they apply, and the way that they apply them. This process, called appropriation, allows them to improve their performance over time. Prior research has, indeed, recognized the role of appropriation in explaining the performance of teams using information technology. The “Fit-Appropriation Model,” (FAM) (Dennis et al, 2001) holds that performance is affected by both technology fit and appropriation. But, until now, there was no recognition of the possibility, much less the likelihood, that appropriation would ever dominate over fit, and certainly not in such a short period of time.
This research is limited to a single context, task, and technology, and used students as research subjects. Generalizability still needs to be established. However, assuming that the findings stand up to further scrutiny, they are ground breaking.
Wednesday, April 8, 2009
Thursday, March 26, 2009
One More Benefit of Good Web Design
Reference: Parboteeah, D.V., Valacich, J.S., and Wells, J.D. (2008) The influence of website characteristics on a consumer's urge to buy impulsively, Information Systems Research 20 (1), 60-78.
It's nice to know that a website's task-relevant cues (ie., appropriate information and good navigation) and mood-relevant cues (ie., visually appeal) increase consumers' urge to buy impulsively. Not that any designer would ever want to create a website deficient in either of these attributes.
While I have some doubts about the relevance of this study, I have little reason to doubt that improving website quality improves outcomes, including impulsive purchasing. The predicted magnitudes are somewhat suspect due to the nature of the experiments. All experiments were performed with students in a classroom setting and in no case were they actually buying anything -- they were simply reporting on their urge to buy. While students might represent the typical demographic of a web purchaser, the setting is anything but typical.
According to the first experiment, which used structural equation modeling, increasing visual appeal by one standard deviation increases urge to buy by .42 standard deviations and increasing information fit to task by one standard deviation increases urge to buy by .29 standard deviations. According to the second study, which used MANOVA to compare sites that had poor task-relevant and mood-relevant clues to sites that had good clues of task-relevance, mood-relevance, or both, impulsiveness increased from 61.1% of participants, to 72.2% with mood-relevant only, to 74.1% with task-relevant only, to 98% with both. Likewise, the magnitude of intended impulse buying increased from $33.89 to $49.17 to $55.56 to $66.39. Of course, this result is heavily dependent on what's being sold and what the potential impulse buys might be. In the experiment, the intended purchase was a $15 cell phone holster and the possible impulse buys were a $60 bag and several $15 accessories.
It's nice to know that a website's task-relevant cues (ie., appropriate information and good navigation) and mood-relevant cues (ie., visually appeal) increase consumers' urge to buy impulsively. Not that any designer would ever want to create a website deficient in either of these attributes.
While I have some doubts about the relevance of this study, I have little reason to doubt that improving website quality improves outcomes, including impulsive purchasing. The predicted magnitudes are somewhat suspect due to the nature of the experiments. All experiments were performed with students in a classroom setting and in no case were they actually buying anything -- they were simply reporting on their urge to buy. While students might represent the typical demographic of a web purchaser, the setting is anything but typical.
According to the first experiment, which used structural equation modeling, increasing visual appeal by one standard deviation increases urge to buy by .42 standard deviations and increasing information fit to task by one standard deviation increases urge to buy by .29 standard deviations. According to the second study, which used MANOVA to compare sites that had poor task-relevant and mood-relevant clues to sites that had good clues of task-relevance, mood-relevance, or both, impulsiveness increased from 61.1% of participants, to 72.2% with mood-relevant only, to 74.1% with task-relevant only, to 98% with both. Likewise, the magnitude of intended impulse buying increased from $33.89 to $49.17 to $55.56 to $66.39. Of course, this result is heavily dependent on what's being sold and what the potential impulse buys might be. In the experiment, the intended purchase was a $15 cell phone holster and the possible impulse buys were a $60 bag and several $15 accessories.
Sunday, March 15, 2009
Yet Another Adoption Model
Reference: Dong, L., Neufeld, D.J., and Higgins, C. (2008). Testing Klein and Sorra's innovation implementation model: An empirical examination, Journal of Engineering and Technology Management 25(4), 237-255.
Despite its title, this article is about the adoption of new information systems, not innovation (except to the extent that new systems can be called innovation). I had previously been unaware of Klein and Sorra’s model (Klein, K.J., Sorra, J.S., 1996, Academy of Management Review), which on its face seems similar to, but not as robust as the UTAUT (Venkatesh et al, 2003, MISQ). The main model finds “implementation effectiveness” to be dependent on “user affective commitment” and “implementation climate.” Implementation climate is, in turn, composed of skills, incentives, and the absence of obstacles. User affective commitment is dependent only on “innovation-values fit.”
To motivate their study, the authors cite the oft-reported research that documents how few companies complete their implementations on time, within budget, and with the promised features and functions. Unfortunately, their model does not address many of the common causes of these failures, such as poor estimates of development costs and time, lack of communication between developers and users, inexperience with the technologies employed, etc. Furthermore, their dependent variable, termed implementation effectiveness, is really just a measure of intention to adopt, as it includes five items that address only the following only the following components: 1) Avoidance, “If I can avoid using the system, I do”; and 2) Endorsement, “I think the system is a waste of time and money for our organization (reverse coded)”.
Contributions of the study include scales to measure implementation climate (5 components, 17 items), innovation values fit (3 components, 13 items), skills (6 items), incentives (2 items), absence of obstacles (3 items), and commitment (4 items). Some of these scales are adaptations from other sources. It is worth noting that the variable “innovation values fit” is similar to the construct of “perceived usefulness” in the TAM and TAM2 models, and to elements of “performance expectancy” and “effort expectancy” in the UTAUT model. It’s components are Fit re Quality, such as “The system maintains data I need to carry out my task,” Fit re Locatibility, such as “The system helps me locate corporate or department data very easily,” and Fit re Flexibility and Cooperation, such as “The system supports the repetitive and predictable work processes.”
The study concludes that “when implementation climate is strong and innovation-values fit is present, an implementation was more likely to succeed than when either climate or fit were weak”.
Despite its title, this article is about the adoption of new information systems, not innovation (except to the extent that new systems can be called innovation). I had previously been unaware of Klein and Sorra’s model (Klein, K.J., Sorra, J.S., 1996, Academy of Management Review), which on its face seems similar to, but not as robust as the UTAUT (Venkatesh et al, 2003, MISQ). The main model finds “implementation effectiveness” to be dependent on “user affective commitment” and “implementation climate.” Implementation climate is, in turn, composed of skills, incentives, and the absence of obstacles. User affective commitment is dependent only on “innovation-values fit.”
To motivate their study, the authors cite the oft-reported research that documents how few companies complete their implementations on time, within budget, and with the promised features and functions. Unfortunately, their model does not address many of the common causes of these failures, such as poor estimates of development costs and time, lack of communication between developers and users, inexperience with the technologies employed, etc. Furthermore, their dependent variable, termed implementation effectiveness, is really just a measure of intention to adopt, as it includes five items that address only the following only the following components: 1) Avoidance, “If I can avoid using the system, I do”; and 2) Endorsement, “I think the system is a waste of time and money for our organization (reverse coded)”.
Contributions of the study include scales to measure implementation climate (5 components, 17 items), innovation values fit (3 components, 13 items), skills (6 items), incentives (2 items), absence of obstacles (3 items), and commitment (4 items). Some of these scales are adaptations from other sources. It is worth noting that the variable “innovation values fit” is similar to the construct of “perceived usefulness” in the TAM and TAM2 models, and to elements of “performance expectancy” and “effort expectancy” in the UTAUT model. It’s components are Fit re Quality, such as “The system maintains data I need to carry out my task,” Fit re Locatibility, such as “The system helps me locate corporate or department data very easily,” and Fit re Flexibility and Cooperation, such as “The system supports the repetitive and predictable work processes.”
The study concludes that “when implementation climate is strong and innovation-values fit is present, an implementation was more likely to succeed than when either climate or fit were weak”.
Friday, March 6, 2009
Adaptation to IT-Induced Change
Reference: Bruque, S., Moyano, J., Eisenberg, J. (2008) Individual Adaptation to IT-Induced Change: The Role of Social Networks. Journal of Management Information Systems 25(3), 177-206.
I was interested in this paper because my current research addresses the role of Web-based social networking on innovation and other organizational outcomes. Existing research on Web-based social networking is quite sparse, probably because the technology is so new. So, when I saw the title of this article, I hoped that it would be relevant to my work. It turns out that this research concerns traditional social networks, not Web-based ones, so its relevance to my own research is not direct. Nevertheless, it seems reasonable to extend the authors' conclusions to individuals' extended (Web-based) networks. Thus, it provides some interesting hypotheses for future research.
Even had there been no relevance to my research, I found this article interesting and refreshing. The highlight for me is to see "adaption to IT-induced change" as the dependent variable rather than the common "adoption of technology." There is a significant difference between adoption and adaption, which the authors describe in some depth. Adoption is a binary variable -- either you adopt or do not. Although one can measure the extent of adoption by counting the number of people in an organization who adopt or fail to adopt, adoption remains binary at the individual level. In practice, many changes force employees to adopt to whatever technology is installed, so the real question is how they adapt to these changes. The authors argue and provide references to support the claim that IT-induced changes are harder to adapt to than most other types of change.
The conclusions of the study are not surprising. Adaptation improves the larger the size of the support network and the greater the strength and density of the informational network. The authors defined these networks to include people outside as well as inside the company. This is a significant departure from most studies and makes me optimistic that the results will extend to Web-based social networks. One disconcerting methodological issue is that subjects were allowed to list only five members of their support network and five members of their informational network. I don't believe that there was any measure of the extent to which these networks overlapped. Of course, Web-based social networks are much larger, although they are probably less "strong" or intense.
A significant contribution of the study is the creation of an instrument to measure individual adaptation to IT-induced change.
I was interested in this paper because my current research addresses the role of Web-based social networking on innovation and other organizational outcomes. Existing research on Web-based social networking is quite sparse, probably because the technology is so new. So, when I saw the title of this article, I hoped that it would be relevant to my work. It turns out that this research concerns traditional social networks, not Web-based ones, so its relevance to my own research is not direct. Nevertheless, it seems reasonable to extend the authors' conclusions to individuals' extended (Web-based) networks. Thus, it provides some interesting hypotheses for future research.
Even had there been no relevance to my research, I found this article interesting and refreshing. The highlight for me is to see "adaption to IT-induced change" as the dependent variable rather than the common "adoption of technology." There is a significant difference between adoption and adaption, which the authors describe in some depth. Adoption is a binary variable -- either you adopt or do not. Although one can measure the extent of adoption by counting the number of people in an organization who adopt or fail to adopt, adoption remains binary at the individual level. In practice, many changes force employees to adopt to whatever technology is installed, so the real question is how they adapt to these changes. The authors argue and provide references to support the claim that IT-induced changes are harder to adapt to than most other types of change.
The conclusions of the study are not surprising. Adaptation improves the larger the size of the support network and the greater the strength and density of the informational network. The authors defined these networks to include people outside as well as inside the company. This is a significant departure from most studies and makes me optimistic that the results will extend to Web-based social networks. One disconcerting methodological issue is that subjects were allowed to list only five members of their support network and five members of their informational network. I don't believe that there was any measure of the extent to which these networks overlapped. Of course, Web-based social networks are much larger, although they are probably less "strong" or intense.
A significant contribution of the study is the creation of an instrument to measure individual adaptation to IT-induced change.
Sunday, February 22, 2009
Predicting 50 Years of Compiler Research -- They Can't Be Serious
Reference: Hall, M., Padua, D., and Pingali, K. (2009). Compiler Research: The Next 50 Years, Communications of the ACM 52:2, 60-67.
I was amused to read the title of the CACM article referenced above. One can't quibble with the tag line -- "research and education in compiler technology is [sic] more important than ever." The article starts out well enough, recounting the past 50 years of compiler advances and noting that in the coming decade, research into compiling for multi-core processing and security and reliability will be major challenges. And it's hard to critique the authors' agenda for the compiler community except that it's rather vanilla and based on current conditions and those easily foreseeable for the near future, such as the need to address parallel architectures. But, it's completely unreasonable to expect that anyone can predict now what our needs will be in 50 years. For example, it seems likely that we will need compilers for quantum computing, yet this possibility is not raised. It's also quite likely, especially if you believe Ray Kurzweil, that by then computers rather than people will be building software, implying an entirely different model for the role of people, if at all, in compiler creation.
I was amused to read the title of the CACM article referenced above. One can't quibble with the tag line -- "research and education in compiler technology is [sic] more important than ever." The article starts out well enough, recounting the past 50 years of compiler advances and noting that in the coming decade, research into compiling for multi-core processing and security and reliability will be major challenges. And it's hard to critique the authors' agenda for the compiler community except that it's rather vanilla and based on current conditions and those easily foreseeable for the near future, such as the need to address parallel architectures. But, it's completely unreasonable to expect that anyone can predict now what our needs will be in 50 years. For example, it seems likely that we will need compilers for quantum computing, yet this possibility is not raised. It's also quite likely, especially if you believe Ray Kurzweil, that by then computers rather than people will be building software, implying an entirely different model for the role of people, if at all, in compiler creation.
Thursday, February 12, 2009
Yet Another TAM Article
Reference: Chin, W.W., Johnson, N, Schwartz, A. (2008), A fast form approach to measuring technology acceptance and other constructs, MIS Quarterly 32:4, 687-703.
I'm sure I'm not the only one who's tired of reading articles about the Technology Acceptance Model (TAM). As noted by the authors, there were 698 citations of TAM by 2003 in the Science Citation Index, and fully 10% of the total publications in the IS field prior to 2003 could be classified as TAM studies. There may have been a bit of a drop off in the percentage of publications addressing TAM since 2003, but it always surprises me that TAM articles continue to be published (often in top journals). How can there be anything new to say about it after all this time?
But, there are always exceptions. Don't let TAM fatigue dissuade you from reading this article. It is less about TAM and more about using semantic differential scales instead of Likert scales for IS research. The authors demonstrate that, at least in this case, semantic differential scales are easier and quicker to use, provide an equal degree of construct validity, and produce similar to identical relationships among the constructs measured. This is inspiring. I've always used Likert scales before, but I will seriously consider semantic differential scales in the future. So, for example, instead of asking users to agree or disagree on a 7 point scale with the statement, "Using the system enhances my effectiveness," I will ask users to select among 7 options ranging from "The system is effective" to "The system is ineffective."
I'm sure I'm not the only one who's tired of reading articles about the Technology Acceptance Model (TAM). As noted by the authors, there were 698 citations of TAM by 2003 in the Science Citation Index, and fully 10% of the total publications in the IS field prior to 2003 could be classified as TAM studies. There may have been a bit of a drop off in the percentage of publications addressing TAM since 2003, but it always surprises me that TAM articles continue to be published (often in top journals). How can there be anything new to say about it after all this time?
But, there are always exceptions. Don't let TAM fatigue dissuade you from reading this article. It is less about TAM and more about using semantic differential scales instead of Likert scales for IS research. The authors demonstrate that, at least in this case, semantic differential scales are easier and quicker to use, provide an equal degree of construct validity, and produce similar to identical relationships among the constructs measured. This is inspiring. I've always used Likert scales before, but I will seriously consider semantic differential scales in the future. So, for example, instead of asking users to agree or disagree on a 7 point scale with the statement, "Using the system enhances my effectiveness," I will ask users to select among 7 options ranging from "The system is effective" to "The system is ineffective."
Monday, February 2, 2009
RFID vs. Bar Coding
Reference: Hozak, K. and Collier, D.A. (2008) RFID as an enabler of improved manufacturing performance, Decision Sciences 39:4, 859-881.
I always enjoy reading articles with counter-intuitive conclusions or conclusions that attempt to dispel commonly accepted truths about an issue. This one concludes that unless processes are changed, RFID fails to provide much of an operational benefit, if any, over bar coding. Attempts to improve mean flow time and the proportion of transactions that are tardy by reducing lot size, a practice enabled by RFID, could actually have the reverse affect. Very interesting! These conclusions, and several related ones, are based solely on a simulation, which may be suspect, as simplifying conditions are always assumed. Nevertheless, anyone considering RFID adoption should read this. The other caveat, and perhaps the more important one, is that the benefits to improved information and reliability are not considered. I've always thought that the information benefits of RFID outweighed all production metric benefits, so I'm not terribly disturbed by the conclusions. But for those who are considering adopting RFID for the production benefits alone, these conclusions should be kept in mind.
I always enjoy reading articles with counter-intuitive conclusions or conclusions that attempt to dispel commonly accepted truths about an issue. This one concludes that unless processes are changed, RFID fails to provide much of an operational benefit, if any, over bar coding. Attempts to improve mean flow time and the proportion of transactions that are tardy by reducing lot size, a practice enabled by RFID, could actually have the reverse affect. Very interesting! These conclusions, and several related ones, are based solely on a simulation, which may be suspect, as simplifying conditions are always assumed. Nevertheless, anyone considering RFID adoption should read this. The other caveat, and perhaps the more important one, is that the benefits to improved information and reliability are not considered. I've always thought that the information benefits of RFID outweighed all production metric benefits, so I'm not terribly disturbed by the conclusions. But for those who are considering adopting RFID for the production benefits alone, these conclusions should be kept in mind.
Subscribe to:
Posts (Atom)
