Between remembering and talking to the dead

The platforms under scrutiny here are considered instances of death-tech (Arnold et al. Reference Arnold, Gibbs, Kohn, Meese and Nansen2017; Biçer and Yıldırım Reference Biçer and Yıldırım2022), and many of the issues arising from them are resonant of studies into ‘digital immortality’ and social networks (Stokes Reference Stokes2015; Savin-Baden et al. Reference Savin-Baden, Burden and Taylor2017; Kania-Lundholm Reference Kania-Lundholm, Holmberg, Jonsson and Palm2019; Kasket Reference Kasket2019; Öhman C and Watson Reference Öhman C and Watson2019; Harrington Reference Harrington2020; Sisto Reference Sisto2020; Bassett Reference Bassett2022) and AI in death (Meese et al. Reference Meese, Nansen, Kohn, Arnold and Gibbs2015; Savin-Baden Reference Savin-Baden2021). Scholars have raised important questions about how the development of synthetic afterlife technologies challenges our understanding of personhood, mortality, and, crucially for us here, memory (Hoskins Reference Hoskins2024). As dialoguing with the dead increasingly happens across multiple modalities – varied data inputs, presentation techniques, and AI-driven responses – it suggests ‘new kinds of social links and symbolic rituals’ surrounding death (Sisto Reference Sisto2020) and potentially potent new forms of ‘conversational pasts’ (Hoskins Reference Hoskins2024) produced through human–machine entanglements. These practices are therefore deeply (and explicitly) mnemonic, reshaping how memory is mediated, sustained, and performed through AI’s architectural and interpretive filters. Our concern is not simply with what AI can do to memory in these processes but with how memory is commodified and emotionally shaped as a result.

If AI technologies increasingly shape how individuals relate to the past, they do so quite particularly in contexts of loss and remembrance, where they mediate access to synthetic narratives of the deceased. In such cases, deathbots function as algorithmic mnemonic agents that not only reproduce traces of the dead but also actively participate in shaping how they are remembered. These systems further embed memory within platform logics (Smit et al. Reference Smit, Jacobsen and Annabell2024a), raising significant epistemological and ethical concerns particularly as memory work is increasingly mediated by infrastructures that are opaque and shaped by affective economies.

In this context, deathbots can be seen as part of a broader algorithmic infrastructure of collective memory (Gensburger and Clavert Reference Gensburger and Clavert2024), whereby AI is increasingly understood as redistributing agency, challenging authorship, and reshaping how the past is constructed and experienced (Volynskaya Reference Volynskaya2024). Memory emerges as co-constructed through real-time interaction (Hoskins Reference Hoskins2024) and probabilistic modelling (Smit et al. Reference Smit, Smits and Merrill2024b), raising in turn ethical concerns about transparency and accountability (Richardson-Walden and Makhortykh Reference Richardson-Walden and Makhortykh2024), especially as AI-generated memory products reflect the socio-technical and economic systems from which they emerge (Gensburger and Clavert Reference Gensburger and Clavert2024). As Schuh (Reference Schuh2024) noted, such artificial memory operates through plausibility rather than provenance, complicating truth claims while making synthetic history increasingly indistinguishable from archival reality. These dynamics are intensified in commercial applications, including digital afterlife services (Meitzler et al. Reference Meitzler, Heesen, Hennig and Ammicht Quinn2024) and emotionally resonant platforms (Joanroy et al. Reference Joanroy, Figenbaum, Frère and Millerand2025), where memory becomes affective and commodified; what Pilkington (Reference Pilkington2024) critiques as a form of ‘myopic memory’ under capitalism. Moreover, the persistence of digital traces – often beyond the control or awareness of users – alters traditional processes of forgetting, raising new challenges for how we grieve, remember, and let go (Hoskins Reference Hoskins2024).

Underpinned by AI architectures, the emergence of synthetic afterlife technologies like deathbots reflects a broader shift in societal experiences of death where ‘symbolic representations’ of the dead become pervasive online (Morse Reference Morse2023), and boundaries between the living and the dead increasingly porous (Stokes Reference Stokes2021), not least of all through the ‘trickery’ encouraged by these platforms as new forms of banal deception underscore our interactions with AI (Natale Reference Natale2021: 129). These ‘technologies of illusion’ (Natale Reference Natale2021) are sustained through what Airoldi (Reference Airoldi2022) termed ‘machine habitus’ whereby algorithmic outputs are culturally shaped performances informed as much by training data and design choices as by social norms. Central to these dynamics are user’s ‘algorithmic imaginaries’ (Bucher Reference Bucher2018) – that is, the emotional and cognitive frameworks through which people attribute agency, autonomy, or emotional presence to algorithmic systems. The simulation of empathy and connection (McStay Reference McStay2023) is one of the most powerful – and ethically fraught – features of (death)bots. In the face of death, interacting with these platforms might also be understood as an example of the ‘algorithmic as if’ – a digitally mediated enactment of desires and relationships that reality can no longer fulfil (Kopelman and Frosh Reference Kopelman and Frosh2025: 2393).

Within this context, Meese et al. (Reference Meese, Nansen, Kohn, Arnold and Gibbs2015) argue that memorialising and remembering have become entangled within the commercial start-up economy, raising fears that platforms will prioritise profitable or ‘consumable’ versions of the deceased, and of memory work. We consider deathbots here to function as platforms in that they are governed by algorithms, interfaces, and affordances designed for continuous interaction and data capture (Smit Reference Smit, Wang and Hoskins2024; Smit et al. Reference Smit, Jacobsen and Annabell2024a). Drawing on Poell et al.’s (Reference Poell, Nieborg and van Dijck2019) definition of ‘platformisation’, we highlight the combination of infrastructural, economic and governance dimensions which ‘reorganise’ cultural practices – including mnemonic ones – around platforms. In the context of memory, the platform society (van Dijck et al. Reference van Dijck, Poell and de Waal2018) refers to a socio-technical formation in which digital platforms not only host but actively structure, generate, and monetise remembrance practices, transforming memory into a data-driven, constantly updating, and commercially mediated process shaped by algorithms. Deathbot services transform memories (and death) into a product, generating capital through what Stanyek and Piekut (Reference Stanyek and Piekut2010) term ‘posthumous performances’. As users engage with these bots, they contribute to emotional data that sustain the digital dead while also reshaping their own identities within platform infrastructures as ‘creatures of the update’ (Chun Reference Chun2016). These commercial synthetic afterlives actively engineer affective experiences as part of the emotional AI platform ecosystem (Bakir and McStay Reference Bakir and McStay2025, np). Designed for affective resonance and continued engagement (Richardson-Walden and Makhortykh Reference Richardson-Walden and Makhortykh2024), memory is crafted not only as interactive and potentially co-created but also as (crucially) profitable (Recuber Reference Recuber2023).

O’Connor and Kasket (Reference O’Connor, Kasket, Machin, Brownlow, Abel and Gilmour2022) reminded us that these necro-technologies (as all technologies) are not neutral; rather, they reflect the beliefs and assumptions of the developers designing them. Systems are coded in ways that are culturally and context dependent, raising additional questions about how we represent the deceased, and the past, and how well we understand peoples’ varied motivations for remembering. In this context, synthetic afterlives have raised significant legislative challenges and ethical concerns (Harbinja Reference Harbinja2022), which are also important in the context of memory wok (individual and collective). These relate in part to how data collection and processing by these platforms intersect with privacy and consent issues, especially given companies must navigate responsibilities to both deceased persons’ digital legacies and grieving family members or users (Elder Reference Elder2020). There are concerns too about how these systems build from and exacerbate structural inequalities, and about the commercial exploitation of the dead (Kneese Reference Kneese2023). Alongside concerns about commercialisation and data extractivism, a broader set of ethical issues has emerged regarding the dignity of both the deceased and living users who interact with synthetic afterlives (Buben Reference Buben2015; Brescó de Luna and Jiménez-Alonso Reference Brescó de Luna and Jiménez-Alonso2024; Fabry and Galvão Reference Fabry and Galvão2024).

Despite the growing body of literature on the intersection of memory and AI and on digital afterlife services, there remains a gap in research focusing on user experience and analysing the platforms that mediate these practices (although some studies examine perspectives of potential users, eg, Morse Reference Morse2023, Galvão V et al. Reference Galvão V, Maciel, Pereira, Gonçalves, Oliveira and Silva2021). This short article starts to fill that gap by presenting an introductory analysis of four platforms, critically exploring their user interfaces and the ways they engage commercially with the emotional dimensions of memory and posthumous sociality.

Memory work with deathbots

In this section, we introduce four platforms that encourage users to ‘story’ and archive a person’s past or to use extant digital data to do so. In order to critically explore each platform, we have carried out a series of socio-technical walkthroughs, using our own data and casting ourselves in the role of ‘users’. In three cases, this meant taking on the role of living persons archiving our own life stories and personas as if for people to interact with after our demise (so, the platform was mimicking us). In the last case study (Seance AI), we interacted with the platform in the role of ‘loved ones’ initiating a séance with a person hybridised from our own extant datasets. As researchers attuned to critiques of digital infrastructures and data power, we naturally have our own reservations about using some of the platforms we analyse in this article and have opted to centre our own data and memories rather than recruit participants to share theirs. This was a decision grounded in our ethics and a desire to do no harm to (possibly grieving) persons. We spent approximately two days working with each platform. For Hereafter AI and Almaya, we contributed 20 video/audio memories each; for Seance AI, we uploaded a series of our own Whatsapp messages and email texts; and for You, Only Virtual, we used those same textual resources, plus a series of lengthy voice notes. Our rationale for this methodological choice – using our own data rather than working with those of research participants – was grounded in our own dubiety about these processes, as well as our desire to better understand the ‘feel of algorithms’ in these contexts (Ruckenstein Reference Ruckenstein2023). For more on the methodology, see Kidd and Nieto McAvoy (Reference Kidd and Nieto2025 forthcoming).

While all four platforms focus on preserving and prompting memories, they do so in distinct ways, leveraging different affordances and vernacular to create representations of the deceased. As we will demonstrate, the platforms’ promotional materials and website/app design are all explicitly mnemonic. This emphasis is clearly reflected in their echoing use of terms like ‘memory’, ‘legacy’, and ‘acts of remembrance’, and where the ‘hereafter’, ‘seances’ and repeated mentions of ‘loss’ invoke themes of continuity, which problematise the memory work that is taking place here.

Each platform uses AI to facilitate interaction with the deceased, but they differ significantly in how they do so, and in how faithfully they reproduce the original person’s data. We offer a comparative table with the main aspects in Table 1.

Table 1.

Platform comparison

Note: Comparative summary of deathbot features. Data from user walkthroughs and document analysis. Revised on 25 June 2025.

Almaya and HereAfter emphasise fidelity to user-generated content. Structured video and audio recordings underscore the platform’s narrative, storied qualities. The focus is on creating and preserving personal archives. Almaya’s engine organises personal video recordings into thematic categories (e.g., family anecdotes, life milestones), enabling loved ones to revisit them through Alma, an AI guide to the deceased’s ‘interactive digital autobiography’. HereAfter similarly archives audio recordings and photographs into categories such as childhood or work, enabling interactions using chatbot frameworks to structure voice-based Q&A dialogues, with the option of using Alexa (Figure 1). In both cases, the AI serves primarily as an organisational and retrieval tool, preserving original voices and narratives without generating new material. In both platforms, the user is also creating a ‘pre-mortem’ digital doppelgänger of themselves, which operates under the notion of an extension of personhood for legacy purposes (Iglesias et al. Reference Iglesias, Earp, Voinea, Kramer, Savulescu and Ferrario2025), in which the ‘autobiography’ (Almaya) or the ‘biography’ (HereAfter) contains meaningful information worth remembering by future generations. Both platforms emphasise the security of the data and the ‘forever’ character of the archive, presumably unaltered.

Figure 1.

Almaya app screenshot, October 2024.

By contrast, Seance AI and You, Only Virtual (YOV) rely heavily on generative AI. Seance AI uses GPT-4 to simulate conversations with the deceased, generating responses from user-provided biographical traits and writing samples. These experiences are deliberately stylised and emotionally evocative. They are grounded in the present and not meant to last (Figure 2).

Figure 2.

Seance AI is ‘Contacting Jen’ in the spirit world. Screenshot September 2024.

YOV synthesises large volumes of personal data – social media posts, SMS, and voice recordings – to build dynamic ‘Versonas’ that evolve over time through machine learning. These avatars can simulate personality traits and generate new interactions, blending real input with algorithmic imagination (and ‘hallucinations’). These platforms allow interactions via both text and audio, making the simulation feel continuous and emotionally present, but also raising questions about authenticity, agency and personhood that we will explore in the next section. Both Seance AI and YOV construct (discursively, through platform affordances and explicitly in the case of YOV) synthetic post-mortem ghosts as underpinned by an understanding of personhood as relational – that is, enabling the ‘continuation of certain relational dynamics’ with the deceased (Iglesias et al. Reference Iglesias, Earp, Voinea, Kramer, Savulescu and Ferrario2025).

These four platforms – commercial technology enterprises – illustrate a variety of practices available within the digital afterlife industry at the current time. Almaya and HereAfter prioritize structured interaction, narrative coherence, and authenticity, functioning primarily as preservation tools underpinned by notions of legacy. Seance AI and YOV, in contrast, embrace imaginative continuity and emergent identity through AI, focusing on simulation and dynamic engagement. These platform design choices reflect distinct philosophies: one focused on curating the past, the other on co-creating emotional futures. In the following section, we examine the mnemo-social implications of these two approaches in more detail.

Deathbots as mnemonic interactants

In this section, we explore the complex relationship between mnemonic agency and affect that the above platforms produce, teasing out what ‘conversational pasts’ (Hoskins Reference Hoskins2024) might look like in the context of commercial deathbots. Given the brevity of this article, we focus here on the tensions between past-oriented archival memory and future-oriented generative memory around the themes of [1] personalisation, algorithms and distributed agency; and [2] emotional resonances.

Agorithms and distributed agency

As noted previously, YOV aims for a high degree of personalisation, requiring a unique dataset for each Versona and encouraging the specification of characteristics using prompts like ‘You are…’ In our study, however, this reliance on pre-set phrases produced a repetitive and awkward communication style, as one walkthrough excerpt highlights:

In allowing users to adjust voice preferences and select life-relevant prompts as they record memories – refining and editing as they go – Hereafter ostensibly offers a high level of content control, but the system’s use of predetermined categories introduces a structure that can limit expressive and ‘mnemonic freedom’ (Recuber Reference Recuber2023: 22), as seen in Figure 3.

Figure 3.

HereAfter screenshot, October 2024. The interface where memories are recorded and archived can be seen, featuring text (in grey) from the automated interviewer, and (in purple) the user making their selections.

Almaya similarly offers personalisation through narrative structure, enabling users to organise memories into chapters within a digital library. Alma, the platform’s AI chatbot, then retrieves content via selected categories, making the communication more akin to an interactive archival exploration than a dialogue. In both cases, the agency of the user is limited by the platform’s affordances as their identity is constructed to be legible to the algorithm, embedding the self within a predictive framework of memory prompts, categories, and interface logic. Conversely, YOV and Seance AI simulate agency by producing avatars that appear to think, remember, or care, giving the platform itself a kind of pseudo-agency, shifting emotional weight onto automated outputs and crafting identities from inferred patterns – writing style, tone, inferred traits – produced by an algorithmic process.

There is a tension here then where personalisation and agency rub up against the homogenising tendencies of these systems (Nieto McAvoy and Kidd Reference Nieto McAvoy and Kidd2024), such as the responses given by YOV in Figure 4, which are often generic and lack specificity. This highlights contradictions between the authentic connection and intimacy claimed by these platforms, the flattening of identities, and the manifold deceptions brought about by generative AI.

Figure 4.

A Versona chat, screenshot October 2024.

In another example, Seance AI stands out in our sample for incorporating contextual data such as date and cause of death into each ‘seance.’ Yet, despite gathering these specifics, it avoided discussing them in conversation, as captured in Figure 5.

Figure 5.

Seance AI conversation. Screenshot September 2024.

The platform’s focus on maintaining an upbeat tone (including the overuse of emojis to talk about the person’s death by drowning) might be a misplaced form of ‘thanatosensitivity’ (Massimi and Charise Reference Massimi and Charise2009). It also suggests a pre-determination which feels less than genuine and remarkably sycophantic by design, particularly in an exchange explicitly framed – however, fictionally – as a postmortem ‘seance’ with the ‘spirit realm’.

More generally, Seance AI did demonstrate sensitivity to data inputs, with tone shifting significantly depending on the source material uploaded. When trained on emails, the bot adopted a formal tone: ‘Hello, it’s your old friend… rather peculiar communicating in this manner isn’t it?’ But with WhatsApp data, the style changed entirely - short sentences, emojis, and exclamation marks. However, the system was unable to move between these differing communicative vernaculars in a way that would have felt more human, and ultimately more sensitive in the potential context of grief.

As can be seen, one important element of personalisation is instinctive communications, with both resonances of a deceased person’s personality and a kind of interior consistency being desirable. Such consistency ensures that the ‘delusions’ being crafted ‘sustain our motivation’ to interact with the system (to borrow from Bortolotti Reference Bortolotti2023), but systems clearly struggle with this at present. Evaluating the promise of intimacy, or at least a bidirectional conversation, our practical encounters revealed how challenging it is to replicate authentic connection in the digital legacy environment, especially where memory is algorithmically reconstructed rather than recalled. YOV showed potential in this regard but suffered from glitches, inconsistent voice tone, and shifts between first and third person, all of which stifled human-like communication.

We would suggest that - certainly at present - there is too much friction in these systems to allow ‘imaginal dialogue’ to take place (Brescó de Luna and Jiménez-Alonso Reference Brescó de Luna and Jiménez-Alonso2024: 107). This problem is further compounded by a lack of social cues which are key to trust in communications (Baym Reference Baym2015). It is not enough for two persons or personas to ‘be together’ in this way, ‘their interiorities must resonate’ too (Bosch et al. Reference Bosch, Fernandez-Borsot, Miró I Comas and Figa Vaello2022). Merely exchanging information does not bring us close enough to the affective and substantive realities of everyday conversation to narrow the ontological gap between humans and artificial companions (Bosch et al. Reference Bosch, Fernandez-Borsot, Miró I Comas and Figa Vaello2022), even beyond the grave. Other research has highlighted the positive (if short term for now) emotional impacts of conversing with social chatbots and the possibilities of developing some sort of individual relationship (see, eg, Weber-Guskar Reference Weber-Guskar, Loh and Loh2022), but this was not our experience. This could be because of the nature of our walkthroughs (created with our own personal data and for research purposes), or because of our own understandings of death and the afterlife. This brings us to the notion of personhood in its different manifestations in relation to the digital ghost, the pre-mortem digital twin and the embodied ‘self’ creating and engaging with these deathbots.

Digital legacy platforms clearly hinge on the eventual absence of the physical body and embrace the complex challenge of capturing human personhood digitally – either as a proxy (standing in for) or as an extension of the self (eg, Papacharissi Reference Papacharissi2010). These digital representations, while mediated, are not entirely abstract or disembodied however; they remain anchored in bodily traces, such as voice data, audiovisual recordings, and images as forms of digital human remains (Kidd and Nieto McAvoy Reference Kidd and Nieto2023). The use of voice data in YOV, for instance, emphasises voice as a powerful marker of identity. Though digitally rendered, voice carries physical qualities – inflections, tone, timbre – that evoke a speaker’s physicality. One of the voices we created was described by its intended end-user as ‘too posh’ and ‘the wrong accent’, revealing how even subtle distortions can trigger emotional responses. Some platforms, like Seance AI and HereAfter, incorporate photographs too, layering the digital persona further, and again reinforcing its connection to an absent body.

Whether digital technologies can support a kind of posthumous personhood remains debated. Stokes (Reference Stokes2021) made the case that being a person is a kind of narrative social construction and that as such, a person can live on in systems of representation and meaning - such as those examined here - after bodily death. Meese et al. (Reference Meese, Nansen, Kohn, Arnold and Gibbs2015) suggested that these systems can ‘materialise new forms of persona that persist’ ontologically beyond death (Meese et al. Reference Meese, Nansen, Kohn, Arnold and Gibbs2015: 415). As suggestive as these are, debates on the notions of personhood in relation to machines and data are far from settled (eg, Chun Reference Chun2016; Cheney-Lippold Reference Cheney-Lippold2017) and further complicated in the face of death (Iglesias et al. Reference Iglesias, Earp, Voinea, Kramer, Savulescu and Ferrario2025).

Using these platforms in their current manifestations prompted us to reflect on what aspects of a person might be considered indispensable when preserving their ‘essence’ in digital form and highlighted the huge gulf between these platforms’ promise and their actual outputs. We have suggested elsewhere that the uncanniness of these synthetic afterlives is reassuring as it enables us to glimpse into the remediation process that meshes data and bodies in new versions of Frankenstein’s monster (Kidd and Nieto McAvoy Reference Kidd and Nieto2023). Rather than ‘symbolic immortality’ (Morse Reference Morse2023: 241), or even ‘posthumous persistence’ (Savin-Baden and Burden Reference Savin-Baden and Burden2019: 101), these platforms stutter towards the creation of an ‘idyllic specter’ (Bao and Zeng Reference Bao and Zeng2024: 5) that remains far out of reach.

Synthetic afterlives as affective infrastructures of memory

These platforms function as affective infrastructures of memory, mediating the encoding, retrieval, and generation of pasts through algorithmic systems. Our walkthroughs allowed us to understand and experience something of these systems’ emotional resonances, which we noted took several forms in and around these practices. First, and perhaps most obviously, the underlying motivation for engaging with these platforms is to preserve memories and maintain an emotional connection with loved ones beyond life. The act of curating and sharing personal stories, even though a digital medium, can be seen as an emotionally driven effort to extend and shape a legacy or a relationship. Well known are these platforms’ commercial motivations for encouraging emotional connections via, for example, the simulation of empathy that is designed to increase engagement (McStay Reference McStay2023).

Second, reflecting on the points above about the ‘disconcerting’ and ‘uncanny’ nature of some of these representations, the potential for emotional discomfort or at least complex emotional responses is apparent (see also, Kidd and Nieto McAvoy Reference Kidd and Nieto2023). YOV is particularly prescient here given users are encountering a digital representation that is attempting to mimic a deceased person’s voice, which evokes a particularly strong and complex mix of emotions, perhaps exacerbating feelings of loss or creating an otherwise unsettling experience. Third, we noted that as users we modified the tone of our contributions (eg, in HereAfter or in Almaya) on the basis of who we imagined might interact with them (our ‘loved ones’). This process of emotional editing emphasises the user’s agency (even if distributed) in shaping the emotional tenor of a digital legacy, and in considering the potential emotional responses of future interactants. It also raises questions about whether the avatar created is really intended as a faithful and interactive reflection of an individual or is instead a carefully curated version of self that prioritises – and preserves in aspic – specific aspects of a personality or stories and ways of telling them, designed by platform’s affordances to be legible to the algorithm. Lastly, as noted above, we found times where the tone of a deathbot or conversational avatar seemed out of step with some of the subjects being discussed (eg, death), evidencing the potential for an emotional disconnect between a system’s design and a user’s expectations. Such disjuncture might be jarring at best, and emotionally deleterious at worst.

Memory, both explicit and implicit, is central to these platforms. Almaya and HereAfter operate on the premise that memories are discrete, containable units worth preserving and sharing. This reminds us that, as Wendy Hui Kyong Chun has argued (Chun Reference Chun2008), digital technologies often conflate memory with storage and position it as something fixed. This supports fantasies of perfect digital recall while masking the inherent volatility and impermanence of technological memory; the bits that do make digital memory human-like (Chun Reference Chun2008). In contrast, platforms like Seance AI and YOV aim not to preserve testimony but to evoke fleeting glimpses of the past, recreating fragments of tone and cadence to echo a person’s communicative presence. As mnemonic tools, however, their outputs often failed to meet their imaginal or functional promise. Interacting with these mnemonic resources was an emotionally unpredictable and complex process; very occasionally because a system would seem to find the perfect story to share or tone to use, more often because the output produced would be jarring or unsettling, and sometimes because it would entirely fabricate a ‘memory’ that we would find problematic or vexatious.

As we proposed at the start of this short article, these platforms demonstrate a tension between archival and generative memory modes then. Almaya and HereAfter lean towards the archival: they structure content for future retrieval, preserving fidelity to the original user’s narrative. In contrast, Seance AI and YOV represent memory as generative and affective, using algorithms to produce plausible, emotionally resonant responses, designed to encourage engagement (McStay Reference McStay2023). All platforms though normalise certain ways of remembering, privileging legacy and continuance, linear narratives, and emotionally responsive interfaces. They also enact not just remembrance and legacy, but new forms of personhood – classified, inferred, and composed by algorithms, where the agency of users is limited and distributed, and the agency of the deathbot is fabricated and ultimately deceptive. As we increasingly interact with algorithmic versions of the self and others, the lines between simulation and authenticity, manipulation and empathy, become ever more fragile.