AI’s Role in Civic Institutions: Beyond Efficiency
Recent developments in artificial intelligence have prompted a critical examination of its potential within civic institutions such as government, education, and public libraries. Unlike private companies driven by profit, civic organizations serve the public good, and their goals extend beyond mere monetary returns. The core mission of institutions like public libraries is to provide resources, foster community engagement, and promote a love of learning for all members of the community, regardless of their ability to pay.
In the private sector, efficiency is often a primary optimization goal, as reducing costs directly increases profits. Conversely, in civic spaces, efficiency becomes meaningful only when it enhances effectiveness – ensuring that more of the institution’s services reach a wider range of constituents. For instance, a library might consider using an AI chatbot to answer patron questions online, freeing up librarians to offer in-person assistance, develop educational programs, or support community initiatives.
Shifting from physical card catalogs to digital search systems exemplifies this efficiency-to-effectiveness pipeline. Patrons can now readily determine if a book is available from the comfort of their homes, saving time and resources. However, a singular focus on efficiency can inadvertently undermine the core aims of these institutions.
The Risks of Over-Reliance on AI
Consider the potential pitfalls of replacing human librarians with AI chatbots, particularly for tasks like homework help. An AI that provides inaccurate information could discourage a child from using the library or seeking educational assistance, leading to a long-term loss of effectiveness. Therefore, the deployment of generative AI solutions must be well-considered and purposeful, rather than implemented solely because of perceived trends.
What might seem like an efficiency gain can ultimately decrease the number of loyal patrons and library visitors, which would constitute a loss of effectiveness. Unintended consequences from attempts to improve efficiency can diminish the library’s ability to provide a universal service.
Sometimes there’s a trade-off between maximizing the return on every dollar and providing reliable, comprehensive services.
AI’s Impact on Efficiency
AI is often discussed as a driver of increased efficiency in the workplace and in institutions, but it’s important to challenge some aspects of this idea. The prevalent theory suggests that the integration of generative AI can boost overall productivity. In the language of economics, AI can enable more work to be done by fewer people in the same amount of time. AI, however, is uniquely suited for certain tasks but lacks the nuanced judgment capabilities of a human for others.
AI’s ability to increase the volume of work completed by fewer people is constrained by the nature of the work itself. For instance, if a library chatbot is tasked with answering simple questions, such as hours of operation, a Retrieval Augmented Generation (RAG) system with an LLM could prove effective. Outside of these limited parameters, it’s critical to establish guardrails and prevent the model from providing potentially incorrect information.
Limitations of AI and Human Oversight
Even with a highly efficient chatbot, there will inevitably be questions that require human intervention. This raises several choices for a library. Should librarians be available for fewer hours, hoping the volume of questions align with their availability? Should patrons be encouraged to contact a reference desk or send an email if the chatbot can’t provide the answers? Or, should they seek answers elsewhere?
The most likely scenario is that the patron seeks an answer from an alternative AI platform. This can cause the institution to lose a patron, who may then receive incorrect information from another source. This illustrates that, in civic environments, efficiency and effectiveness can be at odds.
It is important to emphasize that AI isn’t useless for improving civic organizations. It is critical, however, that public services take the time to think about their goals, and weigh whether the two are compatible.
Labor Conversion
While this example appears simple on its face, eventually this can be expanded, but doing so, in fact, takes work. At any time, it is a question of whether “people do work,” or “models do work”. We need to consider the relationship between “people doing work building AI” and “people doing work providing services to people.”
An essential calculation is determining whether conducting the work directly is more efficient than relying on AI models. Building an AI model offers the advantage of reproducibility, which can lead to greater efficiency over time. However, AI engineering is vastly different from the responsibilities of a reference librarian. In today’s economy, the cost of an AI engineer’s time is considerably higher. In this context, spending the same amount of time working at the reference desk is more cost-effective than employing an AI engineer to develop a more sophisticated AI system. In short, the potential benefits must justify this investment.
We should question the assumption that integrating generative AI guarantees a net gain in efficiency in every situation.
The External Costs of AI
Developing and utilizing AI for tasks doesn’t occur in isolation. The adoption of generative AI tools has environmental and economic costs. Recently, the price for using GPT-4.5, for instance, surged. This includes increased prices for input ($2.50 to $75 per million) and output tokens ($10 to $150 per million) that will be reflected in civic institution budgets. Furthermore, generating 100 words using GPT-4 can consume significant water, not to mention the energy consumption and the use of rare earth minerals in GPUs.
Many civic institutions have global goals that involve improving the world and the lives of citizens in their communities. Concerns for the environment should be carefully considered. Organizations with the goal of having a positive impact should weigh the possibility of incorporating AI more carefully.
We should also consider the potential impact of AI on staffing levels. Some advocate for total reductions in staffing, rather than making existing funds go further. This raises important questions about the reallocation of those funds and how they are used to support community residents. It is important to focus not only on AI, but also on helping the community.
Job Loss and Community Impact
Reducing staffing is not an unqualified good for civic organizations and government, but it needs to be balanced against what other use the money will go to. The primary concern is the workers who may lose their jobs. In the private sector, staff are hired and fired as profit margins change, and the priorities are always for profit. In contrast, civic organizations have community-based priorities. To reduce staffing also means, in a real way, to reduce the economic opportunities available for a range of workers. Jobs give participants in the community the ability to provide for themselves. Providing jobs and supporting the economic well-being of the community is a role that civic institutions play.
When we consider the impact of all of these factors, the loss of any workers also has consequences, such as the loss of patrons. If the focus is on AI and the costs of labor, without thought to the workers, then the entire work structure may be in jeopardy.
Conclusion
Deciding how to incorporate generative AI into civic organizations is not a simple decision, because the purpose of such institutions is fundamentally different from that of for-profit companies. Those who build machine learning solutions in the private sector may see potential applications in government, but they need to be conscious of the significant contextual implications.
Future articles will explore how social science research is utilizing generative AI.
