Author:Shur-Ofry, Michal


The various ways in which robots and AI will affect our future society are at the center of scholarly attention. This Commentary, conversely, concentrates on their possible impact on humanity's past, or more accurately, on the ways societies will remember their joint past. We focus on the emerging use of technologies that combine AI, cutting-edge visualization techniques, and social robots, in order to store and communicate recollections of the past in an interactive human-like manner. We explore the use of these technologies by remembrance institutions and their potential impact on collective memory. Taking a close look at the case study of NDT (New Dimensions in Testimony)--a project that uses 'virtual witnesses' to convey memories from the Holocaust and other mass atrocities--we highlight the significant value, and the potential vulnerabilities, of this new mode of memory construction.

Against this background, we propose a novel concept of memory fiduciaries that can form the basis for a policy framework for robotic collective memory. Drawing on Jack Balkin's concept of 'information fiduciaries' on the one hand, and on studies of collective memory on the other, we explain the nature of and the justifications for memory fiduciaries. We then demonstrate, in broad strokes, the potential implications of this new conceptualization for various questions pertaining to collective memory constructed by AI and robots. By so doing, this Commentary aims to start a conversation on the policies that would allow algorithmic collective memory to fulfill its potential, while minimizing its social costs. On a more general level, it brings to the fore a series of important policy questions pertaining to the intersection of new technologies and intergenerational collective memory.


Pinchas Gutter, a Holocaust survivor, is sitting in a room full of students. "My name is Pinchas Gutter," he begins, "I will answer any questions you might have for me." A boy raises his hand. "How old were you when the War ended?", he asks. "I was between the ages thirteen and fourteen when the War ended. In 1945," Gutter answers. A girl asks: "Do you remember any songs from your youth?" Gutter smiles. "This is a lullaby that my mother used to sing to me, and I still remember it. It's in Polish." Still smiling, he starts singing. His audience is fascinated, only the real Pinchas Gutter is not in the room. The conversation takes place with a virtual Pinchas Gutter--a hologram-like image, backed by sophisticated software. (1) The system integrates advanced display technologies, complicated natural language processing AI, and a database of pre-recorded video interviews conducted with Gutter himself. (2) Their combination allows the 'virtual Gutter' to identify the audience's questions, match the most relevant response from the pre-existing database, and present the answer, as originally delivered by Gutter, in what simulates a human conversational interaction. (3)

The virtual Gutter is part of New Dimensions in Testimony (NDT)--a pioneering project of the USC Shoah Foundation that enables people to have conversations with pre-recorded videos of Holocaust survivors and other witnesses to genocide. (5) In a sense, these virtual witnesses are part of a growing phenomenon of AI-based social robots--robots that are engineered to engage with humans in a social-like manner, exercising learning, communication, and adaptive software capabilities. (6) With the development of machine learning and visualization techniques, the use of AI and social robots is expanding, and their impact on future human lives has become the subject of intense law and policy discussion. (7)

Yet, what is largely missing from this conversation, and what is striking in the case of the virtual Pinchas Gutter, is the use of AI and social robots in a way that affects humanity's past, or more accurately our collective memory of the past. This Commentary uses the NDT project as a starting point for a broader discussion of the policy questions pertaining to the interface of AI, collective memory, and the law. How will societies remember the joint past in an era of virtual memory agents? Should the law regulate the use of AI and robots in ways that affect collective memory, and if so, what would be an appropriate policy framework? These questions are at the center of our inquiry. While our point of departure is the case study of NDT's virtual witnesses, our analysis applies to a broader range of cases where collective memory is mediated through AI-based technologies that possess interactive-communicative skills and perceived human-like authenticity. (8)

Part I of this Commentary begins with a brief introduction of collective memory, a concept that is the subject of burgeoning interdisciplinary literature, yet is still largely new to legal analysis. We briefly explain the notion of collective memory, its social value for the construction of collective and individual identities, and the multiple ways that affect its formation.

Part II takes a closer look at the emerging use of robotic memory agents in the construction of collective memory. Relying on interdisciplinary studies, we show that this new medium carries great promise; it allows for interactions that feel natural, encourages trust and empathy, and can help bridge temporal gaps. It may also be particularly important for overcoming or mitigating 'problems of representation' that exist in cases of genocide or other extreme events. (9) Following this discussion, we proceed to explore potential concerns entailed in this new medium of memory construction, identifying two primary challenges: First, the use of AI-based memory agents inevitably involves editorial choices that may not be transparent to their 'end-users.' While such choices are an unavoidable part of each medium that provides information, the traits of robotic memory agents might make these choices particularly invisible. Secondly, these 'modes of memory' are more susceptible to hacking, manipulation by third parties, and other vulnerabilities in comparison to more traditional media that affect memory construction. These concerns are particularly pronounced since this new medium, by its nature, evokes feelings of trust and reliance on part of its 'users.'

Against this backdrop, Part III of this Commentary explores the potential policy responses to these developments. Relying on socio-cultural studies of collective memory and building on the concept of 'information fiduciaries' developed by Jack Balkin, (10) we introduce a new concept of 'memory fiduciaries.' We explain the nature of and the justifications for memory fiduciaries, and demonstrate, in broad strokes, the potential implications of this new conceptualization for various questions pertaining to collective memory constructed by robots. Our purpose is neither to exhaust the discussion, nor to present a comprehensive 'menu' of legal solutions. Rather we aim to start a policy discussion about the important questions that are at the interface of collective memory, new technologies, and the law.


    The notion of collective memory refers to the joint recollection of the past by societies, communities, nations, and additional groups with elements of a joint identity. (11) Largely attributed to sociologist Maurice Halbwachs, the modern concept of collective memory relies on the understanding that memories are, to an extent, a product of social construction. (12) In other words, our memories are not merely the sum of our own individual experiences, but are constructed in part by the groups to which we belong, be they nations, religious groups, minority groups, kinship networks, or other communities. To illustrate, many of us would say, in everyday parlance, that we remember the first human landing on the moon, the Kennedy assassination, or the Holocaust, although we did not personally experience these events and may not have even been born when they occurred. (13) Thus, the focus of collective memory is not on the cognitive processes of individual memory formation and retrieval, but rather on the social processes and elements that shape our memories as groups. (14)

    During the past few decades, the study of collective memory has rapidly developed into a burgeoning, multi-disciplinary field, integrating insights from sociology, history, anthropology, communication studies, and additional areas. (15) This scholarship recognizes that collective memory forms the connection between groups and their past. (16) It is thus necessary for narrating the life-stories of nations and communities, and constitutes a vital part of their collective identities. (17) Moreover, when the relevant groups are minorities or groups that were exposed to atrocities and persecution, collective memory is perceived as a means of empowerment and restoration. (18) Studies further instruct that collective memory is important for the formation of individual identity as well, since the social and cultural groups of which we are a part deeply influence our sense of self and identity. (19)

    Although often relying on historical accounts, collective memory is not synonymous with history. Since it is a product of social construct, it has subjective and normative dimensions, and can more easily encompass a multiplicity of voices and meanings. (20) Indeed, the same historical event--for example, the atomic bomb on Hiroshima--can play an entirely different role in the collective memory of different groups. (21)

    Relatedly, multiple sources affect the...

To continue reading