ChapterPDF Available
Helena Matute, Miguel A. Vadillo and Pablo Garaizar
Web-Based Experiment Control for Research on Human Learning
4 Feb 2012 01:14
© Springer-Verlag Berlin Heidelberg 2012
Web-Based Experiment Control for Research on Human
Evidence-based eLearning; Methodologies of learning research through the Internet; Online experiments on learning;
Online methods for learning research; Research on e-learning; Software for human learning research; Web-based
research on learning; Web-labs for the study of learning
Web-based experiment control software for research on human learning refers to any type of computer program designed
to control the execution of human learning experiments via the Internet. These programs are designed to test the
predictions of the different theories of learning, and to understand under which conditions learning progresses more
rapidly, more smoothly, or more durably, which conditions allow for better learning and which ones induce more forgetting
and interference. An additional purpose of this type of software is to compare the learning that occurs in the standard,
well-controlled, psychology laboratory to that which occurs through the Internet. In order to fulfill all these objectives, such
software needs to present the learning materials in a relatively homogeneous fashion through different computers and
configurations and needs to allow for the manipulation of that are common in the area of learningindependent variables
such as number of trials, number and type of cues, number and type of outcomes, particular combinations of cues and
outcomes, timing between them, and contexts in which the different phases of the experiment take place. The software
also needs to accurately store the ( of causality, predictive responses, the timingdependent variables subjective judgment
in which the participant’s responses take curve, and so on).place, learning
Theoretical Background
Since the 1980s most of the research that has been conducted on how human learning works, which variables affect the
learning process, and how human learning can be explained (and thus, predicted, controlled, and improved) has been
conducted through personal computers. Research participants (usually college students) were presented with stimuli
whose parameters were controlled by a computer in the laboratory. These parameters included key factors such as
timing, relationship to other stimuli, emotional content, or semantic and instructional value, in addition to more basic
aspects such as size, color, intensity, location, or . The speed and quality of the participants learning wassensory modality
automatically assessed through the registration of the way and time in which they responded to those stimuli. The
development of methodologies of learning research and the corresponding software was normally undertaken by the
psychology researchers themselves, and therefore a huge and heterogeneous variety of experimental software exists
today that can be used in the standard (offline) psychology laboratory. The natural evolution of such programs in the age
of the Internet has resulted in that some of them have been adapted to be run online (see Matute et al. ). The main2007a
advantage of this adaptation is that the Internet has certainly come to facilitate the exchange and homogeneity of
procedures and software among researchers and has multiplied the number (and heterogeneity) of participants that can
take part in each experiment. There are also some potential risks of doing research on learning through the Internet, such
as, for instance, the possibility of some participants repeatedly taking part in the same experiment. However, the potential
risks have been explored and several solutions have been suggested that make the impact of these problems almost
negligible (e.g., Kraut et al. ; Vadillo and Matute ).2004 2009
Just like any other experimental psychology software, these web-based programs designed to conduct research on
learning through the Internet will normally comprise several versions of themselves. The reason for this is that each
version needs to include a different level of an whose effect on the learning process is to beindependent variable
investigated. The program will then execute one or the other version each time that a different user (research participant)
clicks the link to start the experiment. For instance, a program designed to test the effect of massive versus distributed
learning, would comprise at least two different versions that will be assigned at random to each participant. One of these
Helena Matute, Miguel A. Vadillo and Pablo Garaizar
Web-Based Experiment Control for Research on Human Learning
4 Feb 2012 01:14
© Springer-Verlag Berlin Heidelberg 2012
versions would use massive training, with the learning trials being presented in a very rapid succession, one after the
other, while the other version could use much longer intertrial intervals. Of course these variables could also be
manipulated in a within-subjects design, with each participant receiving all conditions in a counterbalanced order.
There are many different research questions, theoretical perspectives, and backgrounds from which web-based research
on human learning is nowadays being conducted. In general, we could say that online research on learning is being
conducted to explore (a) methodological developments (to know whether online experiments are as reliable as laboratory
ones), (b) generality of already known phenomena (to test whether well-known learning effects can also be observed
outside of the laboratory), and (c) novel predictions of learning theories (to run new experiments that could either be run
in the laboratory or through the Internet, in which case the researchers can benefit from the larger samples available in
the web). Therefore, different web-based programs have been designed to investigate online different forms of learning
such as, for instance, probability learning (e.g., Birnbaum, and Wakcher ), (Vadillo et al. ),2002 associative learning 2006
or . As an example, in an experiment designed to study how the of control iscausal learning and illusions of control illusion
acquired in the Internet, participants could be instructed to terminate stimuli that are being presented by the computer.
The termination of the stimuli, however, does not depend on the participants’ responses, but on a preprogrammed
schedule. The result is that participants trying to terminate the stimuli will normally learn an illusory causal relationship
between their behavior and the termination of the stimuli. This is an example of how several learning effects, in this case
an illusion of control, can be equally acquired not only in the laboratory but also through the Internet (Matute et al. ).2007b
Important Scientific Research and Open Questions
One of the hottest current debates to which these programs could contribute by providing the necessary evidence is the
issue of whether learning through the Internet is better, worse, or similar than more traditional forms of learning. Both
professional and lay people’s discussions on these topics are often vehement but the is still scarce.scientific evidence
Indeed, a debate exists on whether these two things are actually comparable (there are many variables that differ
between traditional learning in the classroom and web-based learning at home). Experiments that compare how people
learn through the web and how they do in the classroom using exactly the same materials is one of the comparisons that
is needed, and this is what these programs can best do. The few experiments that have so far been conducted on this
topic are showing, for instance, that using the identical e-Learning or digital learning program in the classroom and
through the Internet produces results that are almost identical in both locations. This has been shown in several simple
tasks and through several different computer programs (Vadillo et al. ). On some occasions,associative learning 2006
however, it has been observed that the same simple program produces slightly better and fasterassociative learning
learning in the classroom, when the instructor is present (though silent) and everything is controlled, as compared to the
Internet, where the participants are possibly exposed to a much greater number of distractions while they perform the
experiment (Vadillo and Matute ).2009
Thus, increasing the variability of these programs and the scope of tasks and domains to which they can be applied is of
course one of the greater challenges for the future. Although associative learning is one of the few areas where these
web-based experimental control programs have initiated their development, their extension to other areas, so that online
and can also be compared in more complex domains, is a must. offline learning Information gathering and Internet
is becoming so common in our generation and the ones to come that understanding the particulars oflearning
Internet-based learning and how this compares to more traditional forms of learning should be a research priority. Indeed,
the popular assumption that because much learning currently takes place online it is necessarily better than more
traditional forms of learning has not yet been supported by evidence. An science will surelyevidence-based learning
require that, in the future, most research tools used to investigate the learning processes are equipped with a web-based
If ethical norms for learning research (and for human research in general) are important, this issue becomes even more
critical when the research is conducted through the Internet. Ethical codes of conduct for scientists conducting research
with human participants require, for instance, that the scientists guarantee the voluntariness, anonymity, and informed
consent of the participants. Whereas this is relatively easy task to accomplish when the experiment is conducted in the
University laboratory or in the school classroom, some additional (and different) measures need to be taken when the
experiment is conducted through the Internet. For instance, providing an informed consent screen that the participant has
Helena Matute, Miguel A. Vadillo and Pablo Garaizar
Web-Based Experiment Control for Research on Human Learning
4 Feb 2012 01:14
© Springer-Verlag Berlin Heidelberg 2012
to accept before starting the experiment does not guarantee that the participant has been informed. Most people would
click on the accept button without even reading the information on the screen. Several good strategies to solve this and
other ethical potential problems have already been discussed (e.g., Kraut et al. ).2004
Another challenge for the future will be adapting these programs to run experiments on machine learning. Whereas
humans have been the only users of this type of experimental software in the psychology laboratory, the fact that these
experiment control programs are now being run on the web suggests that machine learning systems could soon be
accessing these programs (whether in controlled or in uncontrolled ways) and providing their data as if they were human
participants. This could be positive, as it could be used to test the degree of learning in artificial intelligence and machine
learning. This should also allow to test the predictions of our theories of learning as the similarity of machine learning with
human learning and be contrasted (see e.g., ). But it will be important that these programs incorporatelearning algorithms
a means by which both types of data, natural and artificial learning, could be easily discriminated.
Finally, an additional potential problem of current software for Internet-based research is its lack of a broad scope
approach. Most of the available software is based on technologies that do not comply with any standard and that are
proprietary or poorly adapted to heterogeneous execution environments. Ideally, the technologies used for this type of
research should be more flexible and should allow, for example, to compare the results of experiments conducted in
different languages, using the same web applications properly localized, or providing a similar user experience (UX) to
those who access the web experiment from mobile devices with limitations regarding interactivity, connection speed,
computing power, and energy consumption.
Associative Learning
Causal Learning & Illusions of Control
Evidence-Based Learning
e–Learning and Digital Learning
Information Gathering and Internet Learning
Learning Algorithms
Machine Learning
Methodologies of Learning Research: Overview
Probability Learning
Birnbaum, M. H., & Wakcher, S. V. (2002). Web-based experiments controlled by JavaScript: An example from
probability learning. , 189–199.Behavior Research Methods, Instruments, and Computers, 34
Kraut, R., Olson, J., Banaji, M., Bruckman, A., Cohen, J., & Couper, M. (2004). Psychological research online:
Report of board of scientific affairs’ advisory group on the conduct of research on the internet. American
, 105–117.Psychologist, 59
Matute, H., Vadillo, M. A., & Bárcena, R. (2007a). Web-based experiment control software for research and
teaching on human learning. , 689–693.Behavior Research Methods, 39
Matute, H., Vadillo, M. A., Vegas, S., & Blanco, F. (2007b). The illusion of control in Internet users and college
students. , 176–181.CyberPsychology & Behavior, 10
Vadillo, M. A., & Matute, H. (2009). Learning in virtual environments: Some discrepancies between laboratory- and
Internet-based research on associative learning. , 402–406.Computers in Human Behavior, 25
Vadillo, M. A., Bárcena, R., & Matute, H. (2006). The internet as a research tool in the study of associative
learning: An example from overshadowing. , 36–40.Behavioural Processes, 73
Helena Matute, Miguel A. Vadillo and Pablo Garaizar
Web-Based Experiment Control for Research on Human Learning
4 Feb 2012 01:14
© Springer-Verlag Berlin Heidelberg 2012
Web-Based Experiment Control for Research on Human Learning
Helena Matute Departamento de Fundamentos y Métodos de la Psicología, Universidad de
Deusto, Bilbao, Spain
Miguel A.
Vadillo Departamento de Fundamentos y Métodos de la Psicología, Universidad de
Deusto, Bilbao, Spain
Pablo Garaizar DeusTo Tech, Departamento de Telecomunicaciones, Universidad de Deusto,
Bilbao, Spain
DOI: 10.1007/SpringerReference_301903
Part of: Encyclopedia of the Sciences of Learning
Editor: Prof. Norbert M. Seel
PDF created
on: February, 04, 2012 01:14
© Springer-Verlag Berlin Heidelberg 2012
Full-text available
As a consequence of the joint and rapid evolution of the Internet and the social and behavioral sciences during the last two decades, the Internet is becoming one of the best possible psychological laboratories and is being used by scientists from all over the world in more and more productive and interesting ways each day. This chapter uses examples from psychology, while reviewing the most recent Web paradigms, like the Social Web, Semantic Web, and Cloud Computing, and their implications for e-research in the social and behavioral sciences, and tries to anticipate the possibilities offered to social science research-ers by future Internet proposals. The most recent advancements in the architecture of the Web, both from the server and the client-side, are also discussed in relation to behavioral e-research. Given the increasing social nature of the Web, both social scientists and engineers should benefit from knowledge on how the most recent and future Web developments can provide new and creative ways to advance the understanding of the human nature.
Full-text available
JavaScript programs can be used to control Web experiments. This technique is illustrated by an experiment that tested the effects of advice on performance in the classic probability-learning paradigm. Previous research reported that people tested via the Web or in the lab tended to match the probabilities of their responses to the probabilities that those responses would be reinforced. The optimal strategy, however, is to consistently choose the more frequent event; probability matching produces suboptimal performance. We investigated manipulations we reasoned should improve performance. A horse race scenario in which participants predicted the winner in each of a series of races between two horses was compared with an abstract scenario used previously. Ten groups of learners received different amounts of advice, including all combinations of (1) explicit instructions concerning the optimal strategy, (2) explicit instructions concerning a monetary sum to maximize, and (3) accurate information concerning the probabilities of events. The results showed minimal effects of horse race versus abstract scenario. Both advice concerning the optimal strategy and probability information contributed significantly to performance in the task. This paper includes a brief tutorial on JavaScript, explaining with simple examples how to assemble a browser-based experiment.
Full-text available
As the Internet has changed communication, commerce, and the distribution of information, so too it is changing psychological research. Psychologists can observe new or rare phenomena online and can do research on traditional psychological topics more efficiently, enabling them to expand the scale and scope of their research. Yet these opportunities entail risk both to research quality and to human subjects. Internet research is inherently no more risky than traditional observational, survey, or experimental methods. Yet the risks and safeguards against them will differ from those characterizing traditional research and will themselves change over time. This article describes some benefits and challenges of conducting psychological research via the Internet and offers recommendations to both researchers and institutional review boards for dealing with them. ((c) 2004 APA, all rights reserved)
Full-text available
When people try to obtain a desired event and this outcome occurs independently of their behavior, they often think that they are controlling its occurrence. This is known as the illusion of control, and it is the basis for most superstitions and pseudosciences. However, most experiments demonstrating this effect had been conducted many years ago and almost always in the controlled environment of the psychology laboratory and with psychology students as subjects. Here, we explore the generality of this effect and show that it is still today a robust phenomenon that can be observed even in the context of a very simple computer program that users try to control (and believe that they are controlling) over the Internet. Understanding how robust and general this effect is, is a first step towards eradicating irrational and pseudoscientific thinking.
Full-text available
In this article we describe some of the experimental software we have developed for the study of associative human learning and memory. All these programs have the appearance of very simple video games. Some of them use the participants' behavioral responses to certain stimuli during the game as a dependent variable for measuring their learning of the target cue-outcome associations. Some others explicitly ask participants to rate the degree of relationship they perceive between the cues and the outcomes. These programs are implemented in Web pages using JavaScript, which allows their use both in traditional laboratory experiments as well as in Internet-based experiments.
Some published reports have emphasized the similarities between Internet and laboratory research on associative learning processes. However, few of them, if any, studied systematic divergences between both types of research methodologies. In the present experiment, we investigated these divergences using an experimental preparation for the study of associative learning. The results show that discrimination and discrimination-reversal can be obtained both in laboratory and Internet experiments. However, the learning rate was clearly better in the laboratory than in the Internet condition. This result suggests that associative learning experiments performed over the Internet should provide participants with extensive training to assure that asymptotic performance is achieved.
The present study aimed to replicate an associative learning effect, overshadowing, both in the traditional laboratory conditions and over the internet. The experimental task required participants to predict an outcome based on the presence of several cues. When a cue that was always trained together with a second cue was presented on isolation at test, the expectancy of the outcome was impaired, which revealed overshadowing. This experimental task was performed by undergraduate students (N=106) in the laboratory and by a different set of anonymous participants over the internet (N=91). Similar levels of overshadowing were obtained in both locations. These similarities show that web-delivered experiments can be used as a complement of traditional experiments.