Conference Agenda

Predictive AI is extensively used in decision-making in medicine, judiciary system, finance and many other domains. Even if these predictive AI are not used directly in making decisions, they aid humans in decision-making processes. For example, COMPASS, a rate of recidivism predicting AI tool, predicts how likely it is for a defendant to commit a crime. It helps judges decide the defendant’s punishment. Similarly, Generative AI is extensively used to generate text, images, audio, and videos. With the use of generative and predictive AI, a plethora of ethical issues are arising related to bias, privacy and security. While there has been an attempt to minimize bias, fairness has been the most sought-after goal of both predictive and generative AI. Despite bias and fairness being a central concept in morality, they have not been clearly defined. The idea of what fairness should be is pluralistic, which means that there is no single standard of fairness. The problem intensifies when fairness is applied to machine learning algorithms; one standard of fairness often contradicts the other. Clearly, not all standards of fairness can be achieved. Consequently, the problem of bias inevitably arises. It can manifest as unfairness towards the marginalized community in algorithmic decision-making or can manifest as cultural bias in predictive and generative AI, respectively. In a predicament like this, recent literature suggests that we need to re-think how we are looking at fairness, including how fairness has been defined over time, owning to historical injustices and evaluating the goals that one tries to achieve by using fairness. In this paper, I bring forth the conundrum surrounding the fairness debate in AI, which gives rise to allegations that machine learning algorithms are biased. Then, I aim to present a comprehensive study of how fairness is approached according to recent literature, leading to an unavoidable conclusion that adopting fairness in the latter manner eventually makes every AI model contextual not only in the way it performs its task but also in the way fairness as a value is implemented.

While it is tempting to think that technologies as artefacts have an author (Franssen et al., 2024), new technologies rarely emerge from a vacuum. Rather, new and emerging technologies are a summation of prior innovations, a cocktail of science, invention, and discovery. Like the ship of Theseus, elements can be swapped out and reconfigured while the intention of the entity remains the same. And these elements, too, are technologies with their own constellations of elements that can be adapted. What technologies are metaphysically can shift with a swap of a processor, update of a lens, frying of a cable, or runtime reduction of an algorithm.

Technology as an entity is fickle. While “phone” has remained a steady term in the past century, how it works and what it is capable of has shifted. This revolution is in no small part due to a constellation of technologies that formed and shifted until landlines became pocket-sized computers. Even on a micro scale, small updates to processors and shifts in materials allow corporations to release “new” models of laptops, phones, tablets (and more) annually. With the advent of the Internet of Things (IoT), allowing for and encouraging connections between disparate technologies, the breadth of entanglement between technologies is bound to make their capabilities even less concrete.

Phillip Brey (2017) described enabling technologies as “technologies that provide innovation across a range of products, industrial sectors, and social domains,” sharing that “they combine with a large number of other technologies to yield innovative products and services” (Brey 2017). Given the trend to connect technologies, the glorification of historical data collection, and the explosion of machine learning and AI services, this understanding of “enabling technologies” might be too limited. As opportunities for connections grow, so too does the scope of enablement.

As the night sky of technology grows increasingly dense, the opportunities for new Technology Constellations increases. These could form fractals of enablement, infinity loops such as AI learning by reading its own blogs or multiple supposedly privacy preserving tools joining together to form a system of surveillance. A culture of connection through open data and APIs, currently pushed as a new ethical model to encourage innovation, makes it challenging to predict ethical challenges as technologies themselves and what they enable can so readily change.

In our paper, we will use a concrete case of advanced radio frequency sensing to explore the nuances of enabling technologies by considering them as Technology Constellations and uncovering the related issues we need to be prepared for with the expansion of IoT and machine learning.

Work Cited

Brey, P. A. E. (2017). Ethics of Emerging Technologies. In S. O. Hansson (Ed.), The Ethics of Technology: Methods and Approaches (pp. 175-192). (Philosophy, Technology and Society). Rowman & Littlefield International.

Franssen, Maarten, Gert-Jan Lokhorst, and Ibo van de Poel (2024). Philosophy of Technology. In: Edward N. Zalta & Uri Nodelman (eds.), The Stanford Encyclopedia of Philosophy (Fall 2024 Edition), URL = <https://plato.stanford.edu/archives/fall2024/entries/technology/>.