Fubon Center Doctoral Fellow Research
Augmenting or Automating?
Breathing Life into the Uncertain Promise of Artificial Intelligence | Kevin Lee
Augmenting or Automating? Breathing Life into the Uncertain Promise of Artificial Intelligence
Kevin W. Lee is a PhD candidate in organization theory at NYU Stern. His research concerns the dramatic transformations in work and organizing that we have witnessed across today’s economy. He has paid special attention to how people living through these transformations define what is valuable and important, focusing on how this has informed their ability to let go of the past and embrace an uncertain future.
Organizations developing artificial intelligence (AI) have had enormous power and influence over the future we collectively face. While many affected communities – be they organized around race, gender, or class – often have not had a say in AI’s design, one type of community has had a distinctive amount of access to and presence within technology development organizations: occupational communities. That is, members’ experience with and expertise in their communities’ work uniquely qualify them to build AI technologies that can perform it. And while little study has been devoted to exploring how these people relate to and approach developing AI technologies, past scholarship predominantly has been built on the assumption that occupational communities are a source of solidarity, and that members will push back against threats to their community’s work and the value of its craft. The literature thereby suggests that people will work to ensure that the technologies they are developing will not automate, or substitute for, their communities, and instead augment, or complement, them.
However, my investigation of these people surfaced some findings that were puzzling given this theoretical backdrop. I have been following a set of developers within an organization building an AI that composes music, all of whom have primarily identified as members of the occupational community their technology will affect: music composers. To gain an understanding of their lives within their organization – a startup which I anonymized as Reverb – I have been using ethnographic methods: an approach with roots in anthropology, where the researcher gains an understanding of a group’s culture by hanging out with its members and experiencing first-hand their way of life. To gain an understanding of their lives within the broader occupational community, I have been comparing how they have talked about it with how music composers unaffiliated with the company have talked about it in my depth interviews with them, asking primarily about their work, and what is at stake with the advance of technologies like AI. In so doing, I have been able to study how the developers have been navigating any tensions that arise between their memberships in both their organization and their community.
Through such study, I have discovered that Reverb’s developers, though initially acting in ways consistent with what the literature might predict, diverged from these behaviors. At first, they set out to ensure alignment between their organization and community, working to make sure that their AI technology would augment music composers, rather than automate them. To do so, the developers engaged in what I call “reflexive imagining”: they filtered prospective technological futures and features through the lens of what they might want from the technology themselves, assuming that their own background as composers allowed them insight into what their broader community might value and want. Specifically, they appreciated technologies which “collaborated” with them by doing work they did not want to do, allowing them the time to do more valued work instead, and assumed that their community would want the same from the AI they were developing. These beliefs were selectively consistent with what the community actually believed: while not representative of everyone, these beliefs did pull on a set of established and prevalent beliefs on the role that technologies should play in compositional work.
However, and though Reverb’s developers had professed a dedication to augmenting their community, they and their organization eventually shifted toward automating these people, creating some misalignment between what their organization intended the technology to do and what they had felt their community might want. Feeling increasing pressure from investors to produce a return on investment, Reverb discovered that composers were not using their AI, and proposed targeting video content producers instead, therein “competing” with the market for human-composed stock music: a form of music extensively used in the background of videos, and from which some composers derived artistic value. And while this conflicted with the developers’ initial intentions, they justified their organization’s shift as still aligned with what their community wanted. Again engaging in “reflexive imagining,” they devalued stock music, arguing that it was work that neither they nor their community wanted to do, and that automating it would allow them to do more valued work. They thus positioned it in terms consistent with how they had always talked about work their technology could defensibly take over, restoring alignment between their organization’s technology and what their community would want. These beliefs again were selectively consistent with what their community actually believed.
My study has implications for how occupational communities may chart out the future of work, especially when positioned within organizations developing technologies that can threaten their craft. It reintroduces the age-old notion that work may be differentially valued across a community, and that some types of work may be considered more valuable than others, placing this insight in the context of how we make decisions about what to hold onto and let go on the frontier of the future. In particular, my study uncovers how unvalued work may be a source of vulnerability in occupational communities, suggesting that its automation may not elicit the kind of resistance that we have come to expect from occupations facing disruptive technologies. Moreover, my study indicates that we need to look closely at who represents occupational communities within organizations, and at how and to what extent their values are aligned with what the community cherishes. Occupations are “imagined” communities: members relate to them primarily through their own local experiences of them, and their resulting imaginations on what these communities might want, shaped by these experiences, may not be representative of the desires of all their members. Looking closely at what those within organizations value will be crucial to understanding who and what within their communities they will be willing to leave behind: an important part of forecasting what futures they will accommodate and produce.
Organizations developing artificial intelligence (AI) have had enormous power and influence over the future we collectively face. While many affected communities – be they organized around race, gender, or class – often have not had a say in AI’s design, one type of community has had a distinctive amount of access to and presence within technology development organizations: occupational communities. That is, members’ experience with and expertise in their communities’ work uniquely qualify them to build AI technologies that can perform it. And while little study has been devoted to exploring how these people relate to and approach developing AI technologies, past scholarship predominantly has been built on the assumption that occupational communities are a source of solidarity, and that members will push back against threats to their community’s work and the value of its craft. The literature thereby suggests that people will work to ensure that the technologies they are developing will not automate, or substitute for, their communities, and instead augment, or complement, them.
However, my investigation of these people surfaced some findings that were puzzling given this theoretical backdrop. I have been following a set of developers within an organization building an AI that composes music, all of whom have primarily identified as members of the occupational community their technology will affect: music composers. To gain an understanding of their lives within their organization – a startup which I anonymized as Reverb – I have been using ethnographic methods: an approach with roots in anthropology, where the researcher gains an understanding of a group’s culture by hanging out with its members and experiencing first-hand their way of life. To gain an understanding of their lives within the broader occupational community, I have been comparing how they have talked about it with how music composers unaffiliated with the company have talked about it in my depth interviews with them, asking primarily about their work, and what is at stake with the advance of technologies like AI. In so doing, I have been able to study how the developers have been navigating any tensions that arise between their memberships in both their organization and their community.
Through such study, I have discovered that Reverb’s developers, though initially acting in ways consistent with what the literature might predict, diverged from these behaviors. At first, they set out to ensure alignment between their organization and community, working to make sure that their AI technology would augment music composers, rather than automate them. To do so, the developers engaged in what I call “reflexive imagining”: they filtered prospective technological futures and features through the lens of what they might want from the technology themselves, assuming that their own background as composers allowed them insight into what their broader community might value and want. Specifically, they appreciated technologies which “collaborated” with them by doing work they did not want to do, allowing them the time to do more valued work instead, and assumed that their community would want the same from the AI they were developing. These beliefs were selectively consistent with what the community actually believed: while not representative of everyone, these beliefs did pull on a set of established and prevalent beliefs on the role that technologies should play in compositional work.
However, and though Reverb’s developers had professed a dedication to augmenting their community, they and their organization eventually shifted toward automating these people, creating some misalignment between what their organization intended the technology to do and what they had felt their community might want. Feeling increasing pressure from investors to produce a return on investment, Reverb discovered that composers were not using their AI, and proposed targeting video content producers instead, therein “competing” with the market for human-composed stock music: a form of music extensively used in the background of videos, and from which some composers derived artistic value. And while this conflicted with the developers’ initial intentions, they justified their organization’s shift as still aligned with what their community wanted. Again engaging in “reflexive imagining,” they devalued stock music, arguing that it was work that neither they nor their community wanted to do, and that automating it would allow them to do more valued work. They thus positioned it in terms consistent with how they had always talked about work their technology could defensibly take over, restoring alignment between their organization’s technology and what their community would want. These beliefs again were selectively consistent with what their community actually believed.
My study has implications for how occupational communities may chart out the future of work, especially when positioned within organizations developing technologies that can threaten their craft. It reintroduces the age-old notion that work may be differentially valued across a community, and that some types of work may be considered more valuable than others, placing this insight in the context of how we make decisions about what to hold onto and let go on the frontier of the future. In particular, my study uncovers how unvalued work may be a source of vulnerability in occupational communities, suggesting that its automation may not elicit the kind of resistance that we have come to expect from occupations facing disruptive technologies. Moreover, my study indicates that we need to look closely at who represents occupational communities within organizations, and at how and to what extent their values are aligned with what the community cherishes. Occupations are “imagined” communities: members relate to them primarily through their own local experiences of them, and their resulting imaginations on what these communities might want, shaped by these experiences, may not be representative of the desires of all their members. Looking closely at what those within organizations value will be crucial to understanding who and what within their communities they will be willing to leave behind: an important part of forecasting what futures they will accommodate and produce.