1 Introduction
Through big data and highly developed algorithms, artificial intelligence (AI) has gained great importance as an embedded element of digital systems (e.g., Duan et al.
2019; Abdulov
2020) and has changed the functioning of many business models (Strandhagen et al.
2017). In this context, more and more authors point towards the opportunities and risks of AI for achieving the Sustainable Development Goals (SDGs) formulated in the UN Agenda 2030 (e.g., Pedemonte
2020; Vinuesa et al.
2020) and thus highlight the effects on society (Dirican
2015; Kuo and Smith
2018; Kaplan and Haenlein
2019a) or the environment (Khakurel et al.
2018). For example, using a consensus-based expert inquiry process Vinuesa et al. (
2020) conclude that AI can enable the achievement of 134 SDG subgoals but also hinder 59 goals. AI solutions can help reduce pollution, losses in production, or carbon footprint, and help address global warming (PWC
2020; Goralski and Tan
2020). Di Vaio et al. (
2020) emphasize the importance of AI in the context of knowledge management systems for sustainable business models and for achieving the SDGs (in particular SDG 12). However, these authors also point out the risks that may arise from AI. For example, due to globally unequally distributed education and resources, the wealth that can be generated through AI could mainly benefit already wealthy and educated groups or individuals, thus increasing social inequalities. Furthermore, AI technologies are associated with resource depletion (Khakurel et al.
2018) and carbon emissions (Dhar
2020). Considering the opportunities and risks associated with the use of AI in the context of sustainability, it seems fundamentally important to re-define the socio-technical concept of AI (van Wynsberghe
2021) against the backdrop of sustainable development (Duan et al.
2019).
Past research has focused on how AI can simplify human decision making (e.g., Arinze et al.
1997; Kahneman et al.
2016; Schneider and Leyer
2019), support process optimization (e.g., Hoeschl and Barcellos
2006; D’Amico et al.
2019), or foster (sustainable) business models (e.g., Haseeb et al.
2019; Di Vaio et al.
2020), as well as the design of human-machine interaction (e.g., Klumpp
2018; Miller
2018). Less attention has so far been paid to the question of which impact people and their personal values have on the use of AI in the context of sustainable development (Duan et al.
2019). Within a corporate context, values are part of corporate culture. Thus, the impact of corporate culture on the use of AI appears as a major research gap. While there are many varying but no exhaustive definitions of corporate culture (Linnenluecke and Griffiths
2010) we rely on central works (e.g., Schein
2010; Cameron and Quinn
2011) to derive a working definition of corporate culture as
a multi-dimensional and multi-level concept which defines the core values, assumptions, interpretations, and approaches that characterize a company and influence the behavior of its members. With the headline “Culture eats strategy for breakfast,” Eckmann and Klenke (
2021, p. 6) humorously describe the strong influence corporate culture can have on the realization of new corporate strategies. In this vein, various studies emphasize the importance of corporate culture for the actions of a company in the context of sustainable and digital development. For example, following Isensee et al. (
2020), it can be hypothesized that corporate culture influences the use of digital technologies in companies, and thus also the use of AI, in terms of sustainable development. Barredo Arrieta et al. (
2020) and Rakova et al. (
2021) highlight the central role of corporate culture for reinforcing responsible AI. Liu et al. (
2019) use the example of corporate culture in China and Chinese culture in general to describe that culture shapes the approach to knowledge management and AI. Similarly, based on a survey among 300 managers, Gerbert et al. (
2018) identify a particular affinity for AI among the entrepreneurial culture of Chinese pioneers. Despite this initial evidence, there is currently no systematic integration of the relationship between corporate culture and AI. This article aims to close this research gap and therefore addresses the question:
How can corporate culture influence the use of AI in terms of sustainable development?
Methodologically, the answer to this question is based on a bibliometric analysis of the literature from 1990 to 2021. In the first step, a normative definition of AI in terms of sustainable development is developed (referred to as SAI). In the second step, the influence of corporate culture on AI is discussed. The article ends with a conclusion and an outlook.
2 Sustainable artificial intelligence (SAI)
AI can be defined as “the ability of a system to interpret external data correctly, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation” (Kaplan and Haenlein
2019b, p. 17). Given the increased awareness about potential negative effects of an unregulated AI (Bjørlo et al.
2021), Vinuesa et al. (
2020) emphasize that “we are at a critical turning point for the future of AI” (p. 7). While the general definition of AI lacks any normative goal, task description, or application rules, there are more and more attempts of linking the concept of AI with normative ideas, such as those associated with the concept of sustainable development. Responsible AI is concerned with ensuring fairness, model explainability, and accountability (Barredo Arrieta et al.
2020; Rakova et al.
2021). Human-centered AI puts human aspirations, such as human rights, social participation, or environmental protection, at the center of AI design (Shneiderman
2020). For example, the Digital Policy Agenda for the Environment of the Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU
2019) demands a targeted use of AI to tackle environmental problems. Ethical AI introduces principles of transparency, justice and fairness, non-maleficence, responsibility, and privacy to AI design and use (Wright and Schultz
2018; Jobin et al.
2019). The normative concept of sustainable development holds the potential to combine and extend these approaches. It thus appears as a useful starting point for deriving a normative understanding of AI, i.e., SAI, which is basically concerned with exploring AI’s contribution to sustainable development. For example, Lee and Oh (
2020) show that promising AI technologies for sustainable industrial development cover eleven topics, including knowledge representation, machine learning platforms, action recognition, optimization and solving, and identification technology. In applying a technology for social good perspective, Bai et al. (
2021) raise the question how AI could create sustainability opportunities for micro and small enterprises. Respective overarching initiatives aiming to promote a sustainable orientation of AI in corporate practice are increasingly developing and can be joined under the umbrella term of the AI for Social Good movement (e.g., Chui et al.
2018; Tomašev et al.
2020):
-
meta-initiatives: UN AI for Good, AI Commons or Climate Change AI;
-
corporate funding programs: Intel AI for Social Good, Google AI for Social Good, Amazon AWS AI for Good, IBM Science for Social Good, Facebook AI for Social Good, groupelephant—ERP, Microsoft AI for earth;
-
other funding programs: Leonardo DiCaprio Foundation, Ellen MacArthur Foundation, Bill and Melinda Gates Foundation, German Federal Environmental Foundation (DBU);
-
academic programs: UN Global Pulse Network, Data Science for Social Good, DFKI4planet;
-
repositories: Oxford Initiative on AI×SDGs, ITU AI Repository, Plattform Lernende Systeme: KI-Landkarte Deutschland.
Apart from van Wynsberghe (
2021), recent works referring to the concept of SAI fail to provide a holistic definition (e.g., Yun et al.
2016; Dhar
2020; Bjørlo et al.
2021; Dahlin
2021; Fernandez-Aller et al.
2021). For example, the working definition given by Bjørlo et al. (
2021) within the contextual boundaries of online decision-making is based on the Brundtland Report (World Commission on Environment and Development
1987). Therefore, Table
1 provides an overview of four approaches most frequently discussed in literature to define sustainable development (World Commission on Environment and Development
1987; Elkington
1999; Lozano
2008; UN
2015) and the derived deeper, holistic understanding of SAI and its requirements. Therein, AI is considered as a means and an end of corporate sustainable development (van Wynsberghe
2021). Given the complexity of the sustainable development concept, AI is not considered as a system that can define what sustainable development could mean for humans.
Table 1
Approaches to redefine AI in terms of sustainable development
1987 | Sustainable development means meeting the needs of present generations (people; especially the poorest) without compromising the ability of future generations to meet their needs (World Commission on Environment and Development 1987) | 1. SAI contributes to the realization of people’s needs in the present (e.g., individual fulfillment of customers’ needs, especially the poor) without endangering the satisfaction of people’s needs in the future (e.g. through the exploitation of natural resources) |
1997 | According to the original Triple Bottom Line concept, companies have a responsibility to manage their business successfully in economic, social, and ecological terms (Elkington 1999) | 2. SAI enables companies to (simultaneously) meet their economic, social, and ecological responsibilities better or at all |
2008 | When managing the Triple Bottom Line, companies have to also consider the time perspective as a fourth dimension (intergenerational perspective) (Lozano 2008) | 3. SAI enables companies to better understand the interaction of the three sustainability dimensions through time, thus considering short, long-term, and longer-term interactions |
2015 | The guiding principle of the 2030 Agenda pursues the goal of enabling a decent life worldwide while permanently preserving the natural foundations of life. With the 2030 Agenda adopted in 2015, the global community under the umbrella of the United Nations committed to 17 global goals (SDGs) for a better future (UN 2015) | 4. SAI helps people to live a dignified life and to preserve the natural basis of life in the long term. To achieve this, SAI supports the UN’s 17 SDGs |
Due to the current global importance of the SDGs for companies and the integration of the different sustainable development approaches in the 2030 Agenda, this paper links the basic understanding of AI in terms of SAI to SDG achievement. For example, Di Vaio et al. (
2020) highlight the link between AI and SDG 12 (sustainable production and consumption). Accordingly, we refer to SAI when AI contributes to people being able to lead a dignified life and the natural foundations of life being permanently preserved. Consequently, in comparison to AI, SAI presents a means for sustainable development, as it is used to support the SDGs (targeted use) (e.g., Di Vaio et al.
2020; Goralski and Tan
2020; Tomašev et al.
2020; Vinuesa et al.
2020; Fernandez-Aller et al.
2021). In turn, SAI forms an end of sustainable development, as the use of AI should be organized sustainably. This includes assessing and overcoming potential negative impacts and harmful effects for society (Bjørlo et al.
2021) (application rules), such as increased waste through an accelerated use and disposal of technical devices or carbon emissions stemming from energy consumption of computing power (Khakurel et al.
2018).
As an intermediate result, it appears that the normative concept of sustainable development provides useful to redefine AI as SAI, which now includes a targeted use and application rules as normative elements. Following up from this, it is of interest to investigate this relation more deeply within a corporate context. A problem to be faced here is that whether the potentials of AI are exploited to promote corporate sustainable development in the sense of the suggested approaches essentially depends on how the actors in the company use AI and pursue sustainable goals (e.g., sustainable business models) (Di Vaio et al.
2020). In other words, just as companies choose different approaches towards corporate sustainability (Hahn and Scheermesser
2006), they will probably choose different approaches towards AI, which makes the realization of SAI a very complex issue for which no unilateral blueprint can be provided. As corporate behavior will be determined by underlying assumptions and norms as well as other factors related to corporate culture, the following section examines the influence of specific features of sustainability-oriented corporate culture on the use of AI in the sense of SAI.
4 Discussion
The results summarized in Table
3 suggest that specific features of a sustainability-oriented corporate culture influence the use of AI in the sense of SAI (as defined in section 2 and Table
4) in different ways and vice versa.
Table 3
Summary of main findings and propositions
Features of sustainability-oriented corporate culture |
RQ: How can corporate culture influence the use of artificial intelligence in terms of sustainable development? | – |
1 | Attitudes, beliefs, values | Pro-environmental attitude and environmental values of managers and employees | Technology acceptance |
Transparency as corporate value |
2a, b | Behavior (incl. communication) | Proactive pro-environmental behavior of the individual person and the company as a whole | Affection of sustainable behavior |
3 | Collaboration | Willingness and ability to cooperate with internal and external stakeholders | Optimized stakeholder collaboration |
New perspective: human-AI collaboration |
4 | Ethics/norms (incl. leadership) | Accepted environmental responsibility | Increased transparency |
Development of environmental standard |
Sustainable leadership |
AI application rules |
5 | Internal capabilities | Environmental knowledge | Means for increased improved knowledge management and pro-active awareness |
Pro-active environmental awareness |
SAI literacy |
6 | Strategic orientation | Long-term thinking | Optimized forecasting |
Pro-active action |
Targeted AI use |
Other (with a mutual influence on corporate culture) |
7 | Corporate structure | High agility of employees and leadership in work processes | – |
Table 4
SAI as means and ends of sustainable development
SAI as means of sustainable development | AI that is used to support the SDGs and thus contributes to people being able to lead a dignified life and the natural foundations of life being permanently preserved |
SAI as an end of sustainable development | AI use that follows sustainability principles (e.g., carbon neutrality of computing power and life cycle management of AI gadgets, transparent use) |
4.1 Core features of a sustainability-oriented corporate culture
The first partial result of the literature analysis derives the different features of a sustainability-oriented corporate culture from existing conceptualizations. Following the number of conceptualizations considering the features, strategic orientation seems to be of greatest importance. Corporate structure cannot be clearly assigned to corporate culture, although the dependency on a sustainability-oriented corporate culture is to be assessed as high. Overall, we suggest six cultural features that could potentially influence the use of AI in the sense of SAI and are in turn affected by AI.
4.2 Manifestations with an impact on SAI adaption
The results suggest that certain manifestations of the investigated six features of a sustainability-oriented corporate culture are more likely to contribute to the realization of SAI (examples presented in Table
3). For example, a company that holds values related to ethics and transparency (Proposition 1) and has already developed a green code of ethics (Proposition 4) is more likely to internalize SAI ethics and transparency (Khakurel et al.
2018; Jobin et al.
2019; Barredo Arrieta et al.
2020) and thus formulate AI application rules. Proposition 3 supports the manifestation of collaboration, which forms a relevant digital future skill (Kirchherr et al.
2018). Proposition 5 suggests that the higher the awareness and the better the knowledge of sustainability issues and potential applications of SAI as a means for sustainability (SAI literacy), the more likely it is that AI will be used in the sense of SAI.
4.3 Temporal effects
Depending on the feature, the temporal effects of a sustainability-oriented corporate culture on SAI can be distinguished. Some features have an immediate impact (e.g., sustainable behavior), while others have a medium to long-term impact (e.g., ethics). For example, the basic underlying sustainable attitudes and values are unconscious and cannot exert a direct influence on the use of AI in the sense of SAI. Instead, they have a downstream effect by influencing other cultural characteristics, such as behavior. Behavior is more likely to have an immediate and direct impact on the realization of SAI, because the adaption of SAI can be embedded more easily into the structures that have evolved as the means for pro-environmental behaviors, such as responsibilities or work processes.
The results extend previous works suggesting that the use of AI in a company can create the basis for SAI (e.g., Metcalf et al.
2019; Wirtz and Müller
2019), for example in influencing the corporate culture that would facilitate the use of AI in the sense of SAI. Duan et al. (
2019) indicate that the “acceptance and successful application of AI for decision making may result in a change of culture in organizations and in individual behavior” (p. 16). Following Barro and Davenport (
2019), it would be desirable for companies to actively address the human-machine relationship to make the corporate culture smarter and to promote the use of AI in the sense of SAI. In answer to this, Proposition 5 suggests that if AI improves internal capabilities, it is likely that the company’s behavior (including the use of AI) becomes more sustainable in the sense of the SDGs. However, following biases and problems of AI discussed in scholarly literature, AI could also hinder the realization of SAI under certain conditions (Tambe et al.
2019). It can be argued that there is a spectrum ranging from unsustainable AI to SAI. While a detailed discussion on the negative effects of AI is beyond the scope of this paper, the different sections of the sustainability spectrum of AI need to be given equal consideration in the future. Overall, unsustainable AI practices must be avoided, as they would hamper corporate culture from developing in a sustainable manner. We pointed out that unsustainable AI practices could enhance resistance against the decisions and thus reduce AI acceptance among employees or even lead to negative social outcomes, such as anxiety (Moore
2019). A human resource decision, such as the restructuring of workplaces, based on data-driven algorithms that prioritize economic outcomes because participative development approaches including different stakeholders have been neglected would form an unsustainable practice.
Overall, it can be assumed that the more sustainability-oriented the characteristics of a corporate culture, the higher the probability that the company will use AI in the sense of SAI. In turn, it can also be deduced across the board that the use of AI can promote a sustainability-oriented corporate culture (e.g., through increased knowledge management and forecasting).
5 Conclusion
Based on a systematic literature analysis, this paper explores the question in which way sustainability-oriented corporate culture can influence the use of AI in terms of sustainable development (SAI in short). The motivation for this is the potential of AI to influence sustainable development positively or negatively and the lack of a normative guide for AI use. First, this paper contributes to the AI for Social Good debate by providing a normative definition of SAI as a means and end to sustainability with a targeted use and application rules derived from prominent approaches towards sustainable development, including the SDGs. Second, following Barro and Davenport (
2019), this paper demonstrates that the opportunity to use AI in terms of SAI depends on the ability of a company and its corporate culture to innovate AI in terms of sustainable development through its human capital. A first theoretical frame of reference is offered for the two-sided connection between features of sustainability-oriented corporate culture and the use of AI in the sense of SAI. Overall, the presented evidence suggests a strong mutual influence. As practical implications, a variety of starting points can be derived for companies that wish to apply AI in the sense of SAI. Decision-makers can analyze the sustainability-orientation of the corporate culture and check whether the prerequisites for SAI are in place. Therefore, they can draw from the six propositions which systematically demonstrate the potential influence of specific manifestations of cultural features derived from existing concepts of sustainability-oriented corporate culture.
Future research avenues emerge from the propositions and the limitations of this paper. There is a need for further research regarding (i) the conceptual delimitation between SAI, responsible AI, ethical AI, and human-centered AI, (ii) the influence of SAI on corporate culture and corporate structures, (iii) a validation of the effect of the different cultural features on SAI, and (iv) an investigation of the effects of SAI on sustainability. Empirical studies and longitudinal case studies (e.g., good practice cases) would allow for an in-depth investigation of the development phases of SAI, including the identification of cultural manifestations with the most significant and strongest influence based on widely accepted corporate culture concepts and assessment tools.