Paulo Rupino da Cunha (Universidade de Coimbra, Departamento de Engenharia Informática, Pólo II, Pinhal de Marrocos, Coimbra, Portugal)
António Dias de Figueiredo (Universidade de Coimbra, Departamento de Engenharia Informática, Pólo II, Pinhal de Marrocos, Coimbra, Portugal)

The increasing influence of socio-organizational issues in information systems poses serious challenges to the applicability of most traditional research approaches, since the mechanisms from which they derive their rigor and validity become more and more unrealistic in the new contexts.
Indeed, traditional problem decomposition, standardization of procedures, and rigorous quantitative measurement under the control of independent researchers often lose sense in such contexts. We analyze action-research as an alternative approach, bearing in mind that its adoption requires careful consideration of the epistemological foundations that legitimate its use. In particular, we address the legitimacy of generalizing from the researcher's findings, bearing in mind Karl Popper's critical rationalism, to conclude that a virtuous relationship exists between critical rationalism and the mechanisms from which action-research draws its rigor and validity.

We have been witnessing a dramatic evolution in the way information systems support organizations.
Once merely devoted to the automation of unambiguous repetitive tasks, information systems now deeply influence the very business models on which companies rest, affecting all actors at all levels.
Project issues, once confined to the merely technical, are now dominated by complex socio-organizational concerns, making conventional information systems design approaches quite inadequate (Avison & Fitzgerald, 1999; Baskerville, Fitzgerald, Fitzgerald, & Russo, 1995). To complicate matters even further, the research methodologies traditionally used to develop new approaches also became inadequate, since the principles from which they derived their rigor and validity ­ problem decomposition, standardization of procedures and collection of rigorous quantitative measures under the control of independent researchers ­ also ceased to apply in the new contexts. This has brought to the information systems arena what Kurt Lewin had described as "the limitations of studying complex real social events in a laboratory, [and] the artificiality of splitting out single behavioral elements from an integrated system" (Foster, 1972) in (Checkland & Holwell, 1998).
Indeed, it is now increasingly acknowledged that studying new information systems design methodologies, or alterations to existing ones, is impossible from a socio-organizational viewpoint without intervening in the real word to test the techniques (Baskerville, 1999; Baskerville & Wood-Harper, 1996).
A research study under such conditions, and the additional need to take the responsive and flexible posture required to grant that the knowledge gathered in practice informs the developing theory, raises a whole new set of difficulties.
To start with, going out in the field compromises any attempt to exert control over the variables that affect the study ­ a common approach in traditional laboratory experiments ­ since in such rich environments it is not even possible to acknowledge every relevant variable, much less control them on an individual basis.
The usual requirement of researcher independence also becomes compromised, both by the need of direct involvement to provoke the changes to be studied, and by the individual perception of the results, seldom objective or quantifiable.
The ability to repeat experiences as a way to confirm results presented by others, or as a way to try out different alternatives for the same initial conditions, is, almost always, utterly impossible. On one hand, due to the presence of human beings, the research setting is not statically waiting for its governing laws to be "discovered". It is, rather, in a state of permanent (re)construction (Checkland & Holwell, 1998), strongly contrasting with the stability of phenomena studied in the natural sciences, such as those resulting from gravity or magnetism. On the other hand, the very changes provoked by the researcher in such a setting are irreversible, and affect it forever.
Several authors have delved into the topic of finding research approaches capable of answering those challenges, pointing to action-research as a suitable alternative (Avison, Lau, Myers, & Nielsen, 1999; Baskerville, 1999; Baskerville & Wood-Harper, 1996; Checkland & Holwell, 1998; Galliers & Land, 1987; Lau, 1999; Salmela, Lederer, & Reponen, 2000). As (Baskerville & Wood-Harper, 1996) put it, action-research is "one of the few research approaches that we can legitimately employ to study the effects of specific alterations in systems development methodologies in human organizations".
In the remainder of the paper we discuss this research methodology. This is done on the basis of the reflection we have conducted when supporting its use to establish a new information systems development approach presented elsewhere (Cunha & Figueiredo, 2000), (Cunha & Figueiredo, 2001).
In particular, we analyze issues pertaining to the rigor and validity of results and to the ability to generalize from findings obtained through action-research by examining them under the light of Karl Popper's critical rationalism.
In the following section we briefly review the origins and generic characteristics of action-research. In section 3 we analyze the mechanisms used by this methodology to ensure the rigor of the research process and the validity of the results. The ability to generalize from findings is the subject of section 4. Section 5 sums up with some final considerations.