AI’s Implications for Governance

• May 30, 2023
• Artificial Intelligence

In the past few months, there has been rapid growth of generative artificial intelligence (AI) applications online, met with cautious reactions from large-scale, bureaucratic organizations across the globe. As AI continues to evolve and permeate various aspects of our lives, its implications for governance are both profound and far-reaching. From streamlining administrative processes to enhancing public-private collaborations, AI promises to reshape the very foundations of how societies are governed. 

Broadly speaking, AI systems refer to non-human systems that have the capacity to learn from and imitate human’s intelligent behaviors, especially in processing information (UNESCO, 2020). These systems usually demonstrate learning, prediction, and reasoning capacities through trained models.¹ Accompanied by the growth of applications related to Big Data and the Internet of Things (IoT), AI has also become a critical focus in various fields – e-government studies, public management, business management, marketing, and scientific research (Dwivedi et al., 2019). 

For governments of all sizes, AI has become an increasingly familiar term in recent years. The rising interest builds upon the expectation that AI applications may significantly reshape the government's capacities in harnessing complex digital infrastructure, such as the various IoT components as well as the massive amount of data captured from today’s urban environment (Allam & Dhunny, 2019). However, there remains a noticeable research gap. When discussing the relationships between AI and governance, the existing research frequently adopts a rather generic framing of AI rather than specific forms of AI technology² (Sharma et al., 2020; Fatima et al., 2020; Zuiderwijk et al., 2021). 

Additionally, “governance” requires clarification since, just like the term “artificial intelligence,” there is a diversity of definitions. Whether “governance” refers to the government’s capacities, the actions of governmental agencies, or collaborations among public and private sectors is an important distinction that must be made for in-depth discussions. 

The rise of generative pre-training transformers (GPTs) presents a unique opportunity for our society to discuss AI in the realm of governance, in all its definitions. In this short article, we will briefly cover the rapid evolution of GPT models in relation to different conceptual framings of governance, exploring its potential to drive efficiency, empower citizens, and shape the future of public administration.

The Generative Pre-Training Transformer 

As many may have known, the tool ChatGPT was an achievement of a years-long research and training effort. Building upon the transformer architecture (Vaswani et al., 2017), OpenAI introduced the first GPT model in "Improving Language Understanding by Generative Pre-Training" (Radford et al., 2018). Unlike previous natural language processing (NLP) models, which heavily depended on supervised learning from labeled data, GPT employs an unsupervised generative "pre-training" phase to establish parameters, followed by a supervised stage to fine-tune these parameters for a specific task. This approach incorporates both supervised and unsupervised learning methods and drastically decreases the requirement for human oversight and time-consuming manual labeling. Based on the first GPT model, OpenAI upgraded and launched subsequent versions – GPT-2 in February 2019, GPT-3 in June 2020, GPT-3.5 and ChatGPT in November and March 2022, respectively, and GPT-4 in March 2023. Each generation of GPT adopted the previous architecture but acquired significantly more parameters and text data for training. Compared to previous models, the latest GPT model is equipped with an enlarged volume of training data and the capacity to recognize and analyze images. 

The enthusiasm for AI and generative models largely remained among researchers, subject matter experts, and technological institutions until the release of ChatGPT, which immediately instigated mixed feelings toward AI among the general public - surprise, excitement, concern, and fear. The enhanced capacities to understand, process, and respond precisely to human language inputs, ranging from writing short answers to generating sophisticated content, enable the GPT model to reach beyond just an online chatbot. In summary, the single language model brings humans an unprecedentedly intelligent interactive process. However, this is likely to be just the baseline of GPT’s impact on our society. By providing open APIs to third-party access, GPT performs almost as an infrastructure layer for inventing numerous new applications, embedding itself into human society, and stimulating new paradigm shifts in different aspects, which most daily users of GPT might find difficult to even imagine at the moment.

The Multitudes of Governance³

Existing governance studies commonly suggest that the term is frequently used in different ways and has various meanings (Stoker, 1995; Rhodes, 1996). The origin of “governance” can be traced back to the Latin word "gubernare," which means "to direct, rule, guide," and the Greek word "kybernan," which means "to steer or pilot a ship." In medieval English and French contexts, “governance” refers to the action of governing and the art of governing, respectively. Based on the etymology of “governance,” it is possible to derive a lowest common denominator for various understandings. Governance is a mechanism to pilot society (Torfing et al., 2012). The form of mechanism, however, may vary across different contexts.

Briefly speaking, one of the leading perspectives defines governance as the capacity and efficiency of policy implementation and resource management (World Bank, 2007). Similarly, Fukuyama (2013) also considers governance as the government's ability to make and enforce rules and deliver services. The interaction between bureaucrats’ capacity and autonomy is a determining factor for the quality of government; the level of autonomy can be good or bad, depending on the bureaucracy's underlying capacity. 

Conversely, the other group of perspectives expands on the fundamental understanding that governance pertains to a situation where the line between the public and private sectors has become less distinct. For this definition, governance refers to the governing mechanisms that do not rely solely on the exercise of governmental authority.⁴ In other words, governance refers to networks and processes that include public agencies, private actors, and self-organized groups in a civil society. These stakeholders steer society and the economy through collective action and in accordance with some common objectives.

With all these in mind, it becomes possible to envisage how AI-backed tools might change the government’s internal performance and external relationships with non-governmental stakeholders. In our upcoming article, we will delve into specific use cases corresponding with conceptual frameworks of governance.

¹ There is a great diversity of specific technologies that comprise an AI system. For instance, machine learning (including supervised and unsupervised learning), fuzzy logic, natural language processing (NLP), cognitive mapping, and the digital-physical integrations, such as internet-of-things and robotics, computer vision, and autonomous machines and vehicles (Zuiderwijk et al., 2021).

² For instance, it is frequent to encounter a description such that AI can provide assistance in monitoring, simulating, predicting, or optimizing some complex work procedures, be it energy consumption, resource management, resilience planning. There is no doubt that this description is true, but it requires additional specificities for the general audiences to grasp which AI technology might cause what impacts on governance as well as the society at large.

³ Here, I am adopting an oversimplified categorization of various governance definitions. It is for facilitating discussions on the speculated relationships between governance and AI tools like ChatGPT.

⁴ However, it is important to point out that scholars might hold divergent and nuanced emphasis when explaining governance following this strand of thoughts. For instance, some might focus on the network structure between public agencies and private actors (Kooiman and Van Vliet 1993; Rhodes, 1996); some might be more interested in self-organized groups playing a role in shaping civil society (Jessop, 1998); some highlight the processes through which decisions are made while acknowledging particularly the institutional setting in the backdrop (McGinnis, 2011; Torfing et al., 2012).

Bibliography

Allam, Z., Dhunny, Z.A., 2019. On big data, artificial intelligence and smart cities. Cities 89, 80–91. https://doi.org/10.1016/j.cities.2019.01.032

Dwivedi, Y.K., Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T., Duan, Y., Dwivedi, R., Edwards, J., Eirug, A., Galanos, V., Ilavarasan, P.V., Janssen, M., Jones, P., Kar, A.K., Kizgin, H., Kronemann, B., Lal, B., Lucini, B., Medaglia, R., Le Meunier-FitzHugh, K., Le Meunier-FitzHugh, L.C., Misra, S., Mogaji, E., Sharma, S.K., Singh, J.B., Raghavan, V., Raman, R., Rana, N.P., Samothrakis, S., Spencer, J., Tamilmani, K., Tubadji, A., Walton, P., Williams, M.D., 2021. Artificial Intelligence (AI): Multidisciplinary perspectives on emerging challenges, opportunities, and agenda for research, practice and policy. International Journal of Information Management 57, 101994. https://doi.org/10.1016/j.ijinfomgt.2019.08.002

Fatima, S., Desouza, K.C., Dawson, G.S., 2020. National strategic artificial intelligence plans: A multi-dimensional analysis. Economic Analysis and Policy 67, 178–194. https://doi.org/10.1016/j.eap.2020.07.008

Radford, A., Narasimhan, K., Salimans, T., Sutskever, I., 2018. Improving Language Understanding by Generative Pre-Training.

Sharma, G.D., Yadav, A., Chopra, R., 2020. Artificial intelligence and effective governance: A review, critique and research agenda. Sustainable Futures 2, 100004. https://doi.org/10.1016/j.sftr.2019.100004

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., Polosukhin, I., 2017. Attention Is All You Need. https://doi.org/10.48550/arXiv.1706.03762

Zuiderwijk, A., Chen, Y.-C., Salem, F., 2021. Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agenda. Government Information Quarterly 38, 101577. https://doi.org/10.1016/j.giq.2021.101577