Artefacts, routines, and co-production: a pioneering case of artificial intelligence-based health services in Argentina

ABSTRACT

The article presents innovation in artificial intelligence (AI)-based health services as a co-production process between actors and digital artefacts that increase automation and augmentative routine dynamics. The co-production process is analysed from the routine dynamics and contributions of science and technology studies and complexity theory. An in-depth single case of co-production based on AI is analysed to illustrate the innovation process at the micro level. The analysis reveals four findings: (1) innovation in services based on AI-technological solutions emerges from sociotechnical assemblages enacted by actors, artefacts, and routine dynamics; (2) technological solutions based on AI are emergent properties; (3) co-production of technological solutions based on AI are contextualised on situated action and embedded in a cognitive distribution system that leads to automated and augmentative routine dynamics; and (4) the adoption of standardised AI-based technological solutions transforms institutional arrangements. Implications for policymakers and future research agendas are also outlined.

KEYWORDS:

• Innovation
• artificial intelligence
• co-production
• routine dynamics
• sociotechnical
• service

JEL:

• B52
• L8
• O310
• O330

Acknowledgements

Early versions of this article were presented at the GLOBELICS Conference 2021 and other international academic forums (University of Torino, University of Johannesburg). We appreciate the commentaries received there. We also thank Carlota Perez, Cristiano Antonelli, Hernán Thomas, and Erika Kraemer Mbula for comments and suggestions on early versions.

Special thanks to the reviewers of the manuscript. Thanks also to the editors of this special issue.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Some scholars consider theoretical differences between routines and capabilities (Parmigiani and Howard-Grenville 2011; Salvato 2021; Howard-Grenville and Rerup 2016; Baldessarelli, Lazaric, and Pezzoni 2022). According to them, although the capability perspective has roots in the contributions of Nelson and Winter (1982), the practices perspective comes from routine dynamics (Feldman 2000; Feldman and Pentland 2003). We are interested in routine dynamics and how they relate artefacts with innovation. However, we assume that the theoretical perspectives of Nelson and Winter (1982) and Cohen et al. (1996) were the first wave of contributions to the study of routines and artefacts (technologies) concerning innovation processes (D’adderio 2011, 198).

² For a review of the origin and the development of AI field, see Dick (2019).

³ For an in-depth analysis of the differences between classic evolutionary economics and routine dynamics, see Deken and Sele (2021).

⁴ The cognitive dimension of the change of routines reveals micro socio-political problems. The routines as truces are political processes that suppress conflict (Becker et al. 2005). In terms of the social construction of technology, interpretative flexibility and closure are used to explain how conflicts, resistances, and comprehensive truces regarding some technological solutions develop among the relevant social groups (Pinch 2008; Thomas, Becerra, and S Garrido 2017).

⁵ In Section 2 we present the main theorising concepts and propositions in relation to the aggregate dimension..

⁶ A complex system is a modular network and open structure made up of various functional, connected, and decomposable parts (Dopfer, Foster, and Potts 2004). Complexity approaches consider positive feedback processes between organisations and institutional frameworks as determinants of the innovation process (Antonelli 2009; Robert, Yoguel, and Lerena 2017). In our case, innovation can be seen as an emergent property of a complex system in which AI, algorithms, and co-production with humans and non-humans are emphasised.

⁷ As Woolley (2023) evidenced in this special issue, in the field of medical imaging, distributed cognition systems based on artificial intelligence develop federated learning among different actors, even among competitors. Their findings highlight future research agendas linking co-production processes based on AI and federated learning models.

⁸ (B) is MEDICAL-INSTITUTION.