author
Brief introduction
Zhou Li
Professor, School of Journalism and Communication, Central China Normal University
Chen Muen
Master student, School of Journalism and Communication, Central China Normal University
Read the text
Article
I. Introduction
With the blessing of algorithms, artificial intelligence intelligently processes and accurately pushes the memory fragments left by users on the Internet, encodes massive data into an orderly treasure trove of memories, and builds a fascinating "algorithm time machine". In the encounter with the algorithmic time machine, a new form of memory is being formed, that is, AI-based memory. The so-called AI-based memory is not only a technical form of artificial intelligence that aggregates and processes individual memory, but also a media memory generated by the selective presentation of individual data by the algorithm mechanism. The smart album pulls us back to "that year and today", and the platform usage report looks back on the thousands of mountains and rivers we have traveled, and in the face of sudden data past, sometimes we smile, and more often we seem to be caught off guard. With the gradual penetration of artificial intelligence into memory, the most intelligent thinking activity of human beings, what new form will the relationship between technological media and human subjectivity take on, and whether the space of human subjectivity will be reconstructed, which is not only the continuation of the core issue of technological philosophy in contemporary times, but also the deep question that human beings must face in the new technological ecology.
Research on the impact of artificial intelligence on human subjectivity has been carried out at multiple levels. At the social level, artificial intelligence shapes the social environment, changes the social information ecology, makes human society develop into an intelligent society, and reverses the relationship between people and the technological environment. At the industry level, artificial intelligence is changing the career ecology of various industries, prompting relevant practitioners to continuously update their knowledge and skills and maintain the core advantages that robots do not have. At the individual level, AI promotes the transformation of media from instrumental value to personalized value to meet the value needs around human subjectivity. Most of the previous studies have taken "technology substitution theory" as the basic perspective, believing that artificial intelligence may eventually replace human beings themselves by gradually replacing human functions, which has aroused vigilance about artificial intelligence and worries about the future of human beings. On the one hand, this perspective provides a historical framework for reflecting on technological development, and on the other hand, it also falls into a logical cycle and the accumulation of external functions by ignoring the uniqueness and abruptness of the current technological ecology. In order to break through the linear idea of technological evolution, artificial intelligence should not only be regarded as a technology of "data storage", but should be restored to the specific context of human-machine interaction as an actor in knowledge production and consciousness construction. In this sense, the investigation of AI-based memory provides a special channel for in-depth exploration of the impact of AI on the construction of human subjectivity.
Based on the above background, this paper does not focus on the macro investigation of the impact of artificial intelligence from the perspective of technological change and communication ecology, but focuses on its memory construction and micro power practice in daily scenarios. This paper uses diary research and grounded theory to record individual psychological experience through memory activation products pushed by users to algorithms, and investigates the generation mechanism of artificial intelligence into human memory, as well as the strategies adopted by humans to balance the subject's perception in this process, so as to reflect on the reconstruction of human subjectivity under the trend of human-machine symbiosis, and provide spiritual preparation and action plans for the upcoming era of strong artificial intelligence.
2. Literature review
(1) Concept origin: a technical medium as a memory agent
As a memory agent role in the interaction between various fields of society, media memory has been used as an outsourcing of individual memory long before the emergence of mass media, giving memory a more concrete form and elongating the timeline of human memory. In the era of mass media, through the collection, storage, extraction and dissemination of daily information, the media has formed a platform and core of all human memories dominated by the media, making media memory an important part of social memory. Mass media plays a central role in the process of disseminating collective memory, connecting the images and symbols of the present with the past, so that the present and the past of society can be integrated through the continuous narrative of collective memory. The emergence of digital technology has broken the privilege of mass media to store and disseminate information, and electronic media has recorded and extracted human activities to generate digital memory. Digital memory forms the brick and tile cornerstone of the ever-remixing of public and private, local and global, individual and collective in everyday life, and is pervasive in people's lives. Therefore, the initiative of the media as an actor to intervene in human memory is enhanced, and at the same time, it is better used by the memory subject in the process of memory storage and dissemination, which not only reflects the domestication, shaping and manipulation of people by the media as an actor, but also reflects the intention and action of human beings to technologize themselves. Looking back at the development process of media memory, it is not difficult to see that human beings and technological media have always been in a constant interaction, and media memory has evolved from the storage mode, the extraction mode to the intelligent generation mode examined in this paper, and the increasingly powerful memory agent often means the constantly breeding media subjectivity, and it is in this sense that the technological medium not only memorizes the content but also becomes the media memory itself.
AI memory is the reproduction of personal network data by artificial intelligence technology, which is not only a technical form of artificial intelligence to aggregate and process individual memory, but also a media memory based on algorithms. At present, AI-based memory has penetrated into various daily life scenarios, involving various media technology products, including digital, text, pictures, videos, etc., such as the body monitoring data push of smart watches, the annual report push of music apps, the consumption bill push of payment platforms, and the selected photo push of smartphone albums. In terms of technical logic, AI-based memory uses the interaction data between users and artificial intelligence to construct a recommendation algorithm to output personalized integrated memory data, thus surpassing the technical form of digital memory. In terms of construction, unlike the grand narrative of media memory in the past, AI-based memory focuses more on individual memory, so it is endowed with more personal connotations. It is precisely because of the interactivity and individuality of AI-based memory that people often ignore the other aspect of it, that is, AI-based memory is not a simple storage of data, but a selective presentation of an algorithm mechanism. In order to accurately capture and trace the research object, we need to first solve the question of how to present AI-based memory in different communication contexts.
(2) The construction of human subjectivity by media memory
Since ·Maurice Halbwachs put forward the concept of collective memory in the 20s of the 20th century, memory research has shifted from the field of individual psychology to the social and macro cultural construction process, and the research of media memory is the result of exploration along this path. The use of cultural resources and symbolic resources by media to construct media memory is an important framework for initiating, restricting and rewriting the subject's memory, which plays an important role in cultural inheritance and mobilizes individual cultural identity and value consensus. At the same time, the media not only remembers the past, but also "revises" the past according to the needs of reality and the future, and the generated media memory makes the memory individuals as the subject form a convergent emotional cognition, constantly shaping human understanding of historical traditions, and contributing to the pursuit of historical continuity and identity by human beings. Although media memory researchers do not deny that individual memory is an important basis for their research, the technological ecology of mass communication and the bias of research paradigms make individual memory a "missing link" in media memory research, and some researchers even believe that the two are conflicting. Digital memory has greatly improved the extractability of individual media memory, and also made individual memory enter the field of vision of researchers, and this research is more carried out from the perspective of political economy, focusing on the commodification process of digital memory and its impact on the subject's labor. On the one hand, the previous research enlightens us to constantly torture the social construction power of AI-based memory when it examines the construction of human subjectivity, and on the other hand, it also suggests that we should pay attention to the changes in technical logic and memory construction scenarios, break through the "individual dilemma" of traditional media memory, and fully develop the rich texture of media memory.
In order to make up for the lack of explanatory power of traditional media memory research on individual micro-memory practices in intelligent technology ecology, this study introduces the perspective of technology philosophy. Examining the relationship between technological media and human subjects is one of the core issues in the philosophy of technology. Early philosophers of technology took the corporeal orientation of technological media as a starting point, and saw it as an extension of the human subject. Martin Heidegger regards human subjectivity as a prerequisite for the existence of technological media, and believes that the medium as an appliance plays the role of connecting the distance between people and the outside world, and it has the meaning of existence only if it is accessible and trusted by people. Maurice Merleau-Ponty, like Heidegger, sees the media tool as an extension of the body, highlighting the subjectivity of the human being. However, with the development of technological media, the degree of its compatibility with human subjects has been continuously improved, and the interaction between technological media and human subjects has become a new aspect for technical philosophers to explain the relationship between the two. Kittler, F. even proposed that with the phased development of technological media, technology will replace human beings in the position of subjectivity. Although Kittler revised the early media ontology to the ontology of the relationship between man and the media in his later research, the enhancement of technological subjectivity and the dissolution of human subjectivity have become the basic prerequisites for understanding the relationship between the two. Media technology divides people into physiological structures and information technology, and merges with media into operational networks. Bruno Latour goes further along this line of thought, and his actor network theory sees the world as a collection of relationships, and human and non-human subjects are interconnected and influenced by each other as nodes in the network to complete the transformation of resources. From the perspective of the philosophy of technology, all kinds of human bodily functions can be regarded as internal technological existence and connected with external technological media, and the docking mode determines the relationship mode between human beings and technological media.
To sum up, inspired by the social construction paradigm of traditional media memory and the relational interaction paradigm of technology philosophy, we need to ask that as a "simulacrum" or even a "transcendent" of human function, artificial intelligence blurs the boundary between internal technical function and external technological media, so will the AI-based memory driven by it give rise to a new mechanism for the construction of human subjectivity by media memory?
(3) Technical intermediaries and data power practices
Media memory has never been a "technology-neutral" transmission, but the result of technological translation. Technology is scripted before it is used, empowered, integrated with people, and embodied, influencing people's intentions and actions. The relationship between man and the world is always mediated by technology, which means that technology is in the intermediary position between man and the world, which is divided into embodied relationship and explanatory relationship. Embodiment refers to the integration of technology into the human body, and people regard the ability that technology gives to the body as their own ability, and even ignore the existence of technology, and perceive the world as a whole between technology and the body. Explanatory relationships are when people interpret and construct perceptions of the realistic alternatives provided by technology. As a daily companion of people, the media records and preserves brain memory anytime and anywhere in various forms such as text, pictures and videos, and plays a mediating role in the process of memory transmission, making people's dependence on the media more and more enhanced. Media changes the storage path and form of memory, reduces the cost of memory, and then changes people's memory methods and actions.
Researchers of media memory have long pointed out that the interaction between media memory and personal memory seems to be a beautiful picture of interaction and mutual assistance, but there are many hidden risks behind it. Media amnesia triggers a personal memory crisis, and the deletion of media content or the suspension of the website may lead to the loss of part of the individual's memory; The retrieval efficiency of media memory reduces the mobilization of brain memory, resulting in the risk of accelerated forgetting of individual memory. Data managers analyze the social connotation of big data by mining users' digital traces, and disseminate specific ideologies on this basis, which will affect the objectivity of media memory. The above research reveals that artificial intelligence and complex algorithms, as new media memory constructors, not only reconstruct human memory in terms of technical form, but also may form a new media ideology and compress the space of human subjectivity. However, it is not easy to effectively reflect on the subjective risks of AI-based memory. As mentioned above, AI-based memory is different from the macroscopic construction and strong output of traditional media memory, and it is characterized by the integration of individual memory and human-computer interaction, and has a more hidden construction strategy.
3. Research design
(1) Research methods and data collection
In this study, the diary method and grounded theory were used to collect data, analyze texts, and construct theories. A diary study is a qualitative research method used to collect data on user behavior, activity, and experience, which is recorded by participants over an ongoing period of time. Compared with qualitative research methods such as in-depth interviews and focus groups, the diary method can provide participants with a more free and relaxed narrative space to obtain richer and more authentic empirical materials, so as to help researchers accurately grasp the real-time behavioral and emotional characteristics of the subjects. Our research aims to investigate the memory construction and power practice of artificial intelligence at the individual level, and memory contains a large number of perceptual experiences of individuals, so the diary method can provide more insight into individuals' media use behaviors and emotional experiences. Based on the fact that our research subjects need to have strong activity and sensitivity to platforms using algorithmic mechanisms, 91 eligible college students were recruited in this study. During the two weeks from December 20, 2021 to January 2, 2022, 91 participants were required to record their feelings in a timely manner after each detection of the memory-activated content pushed by the algorithm, and finally a total of 503 valid diaries with a total of more than 160,000 words were recovered. In this study, the qualitative software Nvivo11 was used to encode the recovered valid diaries, and the content of the diaries was analyzed by text analysis.
(2) Coding analysis based on grounded theory
1. Open coding
First, the diaries of 80 participants were randomly selected, and they were split sentence by sentence, read repeatedly, and the expressions with similar meaning were summarized and coded, and a total of 41 initial concepts were obtained after excluding the initial concepts with a frequency of less than 3 (Table 1 and Table 2). Second, these initial concepts were combined into 19 separate categories (Table 2). Finally, in order to test the theoretical saturation, the diaries of the remaining 11 participants were processed in the above way to obtain categories, and compared with the 19 categories sorted out before. The comparative results show that no new concepts are generated, indicating that the concepts and categories summarized in this study have basically reached theoretical saturation.
2. Spindle coding
In this study, 19 categories were compared and classified, and 9 main categories were extracted, namely smart wearable scenarios, knowledge acquisition scenarios, mobile social scenes, lighting up memory glimmers, adjusting memory biases, connecting parallel memories, reflecting on subject illusions, reconstructing memory constellations, and creating memory seals (Table 2).
3. Selective coding
After three levels of coding, the relationships and levels between the nine main categories can be sorted out, and the core categories of this study can be established (Fig. 1). Smart wearable scenarios, knowledge acquisition scenarios and mobile social scenarios are the main practice scenarios of AI-based memory, lighting up the memory glimmer, adjusting the memory bias, and connecting parallel memories are the technical chimeric mechanisms of AI-based memory on human subjects, and reflecting on the subject's illusion, reconstructing the memory constellation and creating memory seals are the reflection and escape processes of users in the face of AI-based memories invading personal memory.
Fourth, the practice scenario of AI-based memory
(1) Constructing a technical body: AI-based memory in smart wearable scenarios
Users wear smart wearable devices all day long, adjust their physical activities according to the real-time state data of their bodies, incorporate smart wearable devices into their body structure, and incorporate their quantitative data into their perceptual experience, thus forming AI-based embodied memory. It is in this process that AI-based memory enables the human body to realize the transformation from a technical body to a technological body. Smart wearable devices use artificial intelligence technology to quantify and feed back to the individual, and the formed AI memory will participate in the person's physical activities, extend the functions of the human body, and make the body technological, and the diary keeper sees this feedback as a clear monitoring of the physical condition. As AI-based memory amplifies bodily senses through real-time monitored data prompts, individuals can concretely feel the abstract physical state of the past, which further prompts the body to continuously adjust to the technical indicators in the frequent interaction with technology, and the technology has also moved from participating in physical activities to becoming a part of the body. In the continuous cycle of man and media, body and technology, man has become a technological existence.
I usually mainly use the (smart) bracelet to check the amount of exercise, sleep time, heart rate changes of the day, especially my heart rate changes a day, so I set the page showing the heart rate fluctuation 24 hours a day as the Home, it can also monitor the blood oxygen index, I occasionally feel hypoxic when I measure blood oxygen often shows that the blood oxygen index is about 90%. Regarding the measurement of sleep and exercise, the comparison is consistent with the facts, and it is more effective for me to adjust my sleep, exercise amount, and monitor my physical condition. (Recorder 5, 12/20/2021 16:03)
(2) Deepening cognitive interaction: AI-based memory in knowledge acquisition scenarios
Compared with the smart wearable scenario, the knowledge acquisition scenario transfers the role of artificial intelligence on the body to individual cognition, so that the AI-based memory is both representational and non-representational. On the one hand, AI-based memory represents knowledge objects as text, pictures, sounds and other symbols embedded in the learner's knowledge system, and displays the user's learning results with indicators such as learning duration, field, and progress. On the other hand, it is detached from the object of knowledge, and the acquisition of knowledge is linked to the subject factors such as situation, individual emotion, and behavioral habits, so as to have non-representation. As a result, AI-based memory has become a highly contextualized knowledge representation, and learners' cognitive interaction ability can be improved by coordinating and integrating the two cognitive frameworks of representation and non-representation. AI memory not only uses artificial intelligence for data modeling, generates representation symbols into intelligent text to assist users to participate in cognitive activities through various structures such as body and brain, but also promotes the interaction between the elements of the internal system of the cognitive subject and the elements of the external system by constructing the corresponding spatio-temporal context. When talking about related scenarios, a large number of diary keepers used words related to positive action, such as "planning" and "persistence", believing that AI-based memory can convert short-term memory into long-term memory and improve learning habits.
The WeChat reading report perfectly reflects my major and hobbies, as well as my studies over the past year, and I feel very focused when I think about it now. In the novel, I find Chekhov's "The Sixth Ward" the most interesting, because each one is not long, so I often read it in half an hour between two classes. The overall feeling for the year is that I still have a lot of room for improvement and should spend more time reading a wider range of books. (Recorder 86, December 30, 2021 17:46)
(3) Presenting data mirroring: AI-based memory in mobile social scenarios
The algorithm logic of social media to present AI-based memories is to sort out historical data from users' social media accounts and highlight a node as a memorable special moment, which has become an "easter egg-shaped function" on many social platforms. In the process of polishing the selected memory into a commemorative moment, the algorithm presents the user's "data-based self" based on the user's network behavior trajectory, and the mirroring of personal memory is also completed. The function of presenting the objective self as an intermediary medium has been around for a long time, and the digital self shaped by artificial intelligence uses individual memory as the material and is doped with algorithmic awareness, so that users can have a more realistic sense of situation and trigger their nostalgia. "QQ Space's 'Today That Year' gave me a review of the content of the third year of high school, a little nostalgic for the past, just care about how many points I got in the monthly exam, and I just had to go to class and brush up on the questions every day. Personally, I think it is a true reflection of my memory. "The deeper impact of AI-based memory on presenting the objective self lies in the fact that it can screen and assign special meaning to the facts that have happened, so that people can recapture the past scenes in the symbolic role of the media and self-imagination, and reshape the subjective self according to the objective self presented by AI-based memory.
These data (presented in the bilibili annual report) provide me with a channel to better understand myself, for example, this year I started to study makeup seriously, so I watched a lot of beauty videos. At the same time, it also sounded a wake-up call for me to reflect on whether I should not spend too much time on entertainment and pastimes, but also invest time in expanding my knowledge to better enrich myself. (Recorder 20, December 31, 2021 10:41)
From the diary text, it can be seen that although users can identify AI-based memories in a single scene, the current integrated memories after AI aggregates and processes individual memories are "pervasive" into a variety of life scenarios. This also makes it more and more difficult for individuals to keep a "distance" from AI-based memories at all times, and the "hands-on state" of human use of media has magically disappeared, and the existence of technology and human existence have become one, or in other words, by integrating human instincts, technology has become a kind of being. The question that arises is, how did all this happen? How does AI-based memory realize the chimerism of individual memory in a deeper memory mechanism?
5. The technical chimerism of AI-based memory to individual memory
(1) Light up the glimmer of memory
Memory glimmer is an individual memory that is outside of power and ignored by the grand narrative. Artificial intelligence can integrate, classify, and polish memory glimmers, and delicately package and push out individual memories that are difficult to tell the public with the help of labels such as time and place. Some researchers have summarized the process of memory travel between the mind and the media as "psychological time travel", in which memories are transmitted and traveled through various media, and are transformed and reconstructed in different time and space. In this sense, AI-based memory opens a psychological time travel of man-machine co-selection, in which individuals store memory glimmers through media, and algorithms gather and select memory materials from them and assign them meaning.
Visualizing and reproducing past scenes is an important way for AI-based memory to light up the glimmer of memory. The visual presentation of the pictures' visual representation of people, places and other elements clears the fog of memory and brings the diary recorder into the memories, "These photos (from the mobile phone album 'That Year and Today') seem to bring me back to last winter, when I ran with my good friends from Jianghan Road to Han Street to Lihuangpi Road, until the gate of the school rushed back to the dormitory at the time of the curfew, that time was precious and beautiful for me." In addition, AI memory can also light up the memory glimmer with the computational integration of network footprints, and many diary keepers have successfully recalled the specific scene at that time with numbers alone, "In 2021, I ordered the most expensive takeaway, which cost 290 yuan, (Meituan's annual bill) I remembered that I went to Nanchang to take the exam at that time, and because I got up late, everyone ordered takeout while washing, which solved the urgent need." Every time a memory is re-spoken in the media space, it is not only the reawakening of the memory by the media, but also the re-superposition of the memory by the media, so that the activity and solidification of the media memory maintain just the right tension. The psychological time travel constructed by AI memory can light up the originally bleak memories of individuals, and also make the originally complex individual memories more intertwined.
Media memory is not only a mental time travel back in time, but also a mental simulation and rehearsal of what might happen in the future, resulting in forward-looking memory. Prospective memory projects the self into the future through the storage, extraction, and dissemination of media memory, so as to experience a psychological time travel to the future. On the one hand, AI-based memories can urge people to stick to past plans, "(Scallop English Annual Report) can bring back memories for me, like, '2021-07-01 05:28, Morningside, you clocked in at Scallops, do you remember that day?'" I remember having insomnia that day, and I started memorizing words at 5 o'clock. This report motivates me to keep learning English." On the other hand, AI-based memories can motivate people to start new projects, "(AutoNavi Map's annual travel report shows) In one year, I lit up two cities, one is Wuhan, unfortunately, and the other is my home. And I only explored 1.7% in Wuhan, reminding me that I must go out more in the new year."
(2) Adjusting for memory bias
The risk caused by traditional media memory bias still exists in AI-based memory, but AI-based memory gives individuals a more active position in the process of adapting memory bias, leaving a gap for individual psychological intervention outside the cultural construction paradigm of traditional media memory. At the same time, the subject's adjustment of memory bias also provides improvement rules for artificial intelligence to further calibrate its matching degree with human memory, and artificial intelligence and human beings complete memory synthesis in two-way running.
The algorithm will inevitably make mistakes in the process of grabbing, calculating and analyzing data, and the algorithm rules are weak, which will distort part of the memory and be different from the human brain memory, "I don't know what kind of statistical method station B is, which leads to statistics that I have watched 59 live broadcasts, but in fact, I have no interest in live broadcasts." For this kind of memory distortion, diary keepers often use empirical calibration strategies to constantly adjust individual memories in the speculation of algorithm rules, "The footprint function of AutoNavi Map reflects part of the user's itinerary throughout the year, which can mobilize my geographical memory of holiday travel and weekend play." The records of some cities and corners are not accurate, perhaps the software is based on the frequency of stops and round-trips." At the same time, artificial intelligence will also patch algorithm loopholes to improve the matching degree with the subject's memory, "Weibo's year-end summary feels very real, presenting the common sentences and expressions that I usually surf and speak."
Personal memory data is stored scattered across platforms, so the AI-based memories viewed by users on a single platform are often incomplete and reduced. Recorder 39 keenly captured this incompleteness, "Xiaomi's smart album is very incomplete, because I went to Suzhou on May Day to play and took a lot of photos, so there were a lot of photos at that time, but it couldn't summarize my life in a year." Compared with memory distortion, the algorithm rules behind memory reduction are more hidden, which also inversely stimulates diary recorders to be more active in adapting behavior, and they often consciously expand the memory bank to compensate for the forgetting of incomplete AI-based memories, "My mobile phone album annual review counts my annual photos this year, this algorithm is really inaccurate, and if you really want to choose an annual review, you still have to sort it out yourself." The algorithm also calibrates the incomplete calculation rules through the data generated by users, and many diary keepers affirmed the memory integration content pushed by the most commonly used social platforms, "The data given by QQ Music's annual listening report presents my listening experience in the past year quite comprehensively."
The memory bias of platform self-interest can be accompanied by false recalls, often in the form of making up an event that never happened. Recorder 63 is aware of this danger, "'The road traveled hides the insight of life, and firewood, rice, oil, and salt are the true meaning of life. 'Ordering takeout is not good for our health in our opinion, but it is so poetic that Meituan said." "Meituan's annual bill shows the total amount I spent and the amount of money Meituan saved me, and in my opinion, merchants tend to raise the price and then discount or issue coupons to create the illusion of price reduction and discount." At the same time, the algorithm will also accurately analyze the user's behavior habits to continuously eliminate false memories and bridge memory biases, "Taobao bills are very real, it determines my purchase habits that make decisions quickly and put cost performance first through the time and choice of my browsing of similar products, I am right." Although the strategies of users and artificial intelligence to adjust memory bias are different, in the cooperation of the former to recognize bias and the latter to correct bias, the conflict between memory and forgetting, technological consciousness and human subject is gradually resolved.
(3) Connecting parallel memory worlds
The virtual world built by artificial intelligence can break the time and space barriers of people's communication and create a parallel world of memory sharing. From the perspective of time, AI-based memory reverses people's perception of time, and taking an algorithmic time machine can make people stop at any time node, breaking the continuity of time. At the spatial level, AI memory can enable people to move in the superimposed world of reality and virtuality, realize people's simulation participation in the virtual world, and subvert the presence of traditional space. Therefore, the object of AI-based memory is not only limited to the individual's record of what happened in the real world, but also includes the interaction between people and AI, and between people in the online virtual world, so that people have parallel memories of the real world and the virtual world at the same time. Just as ·Andrew Hoskins proposed the "connectivity turn", the great abundance and penetration of digital technology and media has made people in the real world actively join or passively participate in the virtual world at any time, forcing people's thoughts and behaviors to flow between the parallel world and the real world, and then reshaping time, space and memory. AI-based memory smooths the boundary between the two memory worlds and makes the flow of memory subjects more free.
In the real world, the interactive performance of the memories of the virtual world is a common means for AI-based memories to forcibly connect parallel memory worlds. By packaging the virtual world memory into a vivid and interesting page design, the AI memory successfully stimulates the user's interactive participation behavior, so that they can't help but immerse themselves in the parallel world created by the algorithm. For example, 'click to start', 'click to pass', 'click to open the door', 'high-five me', 'open the total bill', etc. On the one hand, it can increase the user's sense of participation, and on the other hand, there is a feeling of opening the 'blind box', which can attract users to continue to look down."
Compared with interactive performances, the process of creating twin memories is more natural and ingenious. The memories of the real world and the AI-based memories of the parallel world are intertwined and mapped, and they are mutually constructed into twin memories. By extracting the same feature elements and relationship networks in the two worlds, the AI-based memory realizes the mutual construction of users' memories between parallel worlds, "(in the Honor of Kings Weekly) Seeing these data, I remembered the experience of playing Honor of Kings in the past week, and also recalled what heroes I played. What's more important is the experience I have with my friends and sharing this data with them in our lives makes us all happy."
With the transformation of algorithms from personalization to humanization, algorithms can meet people's psychological needs through "value calculation" and realize users' value recognition of AI-based memory. Compared with the first two levels, "value calculation" transcends the memory connection at the physical level and becomes a value scale based on the spiritual level to connect with the real world. In the view of the diary keeper, the algorithm is not a cold technical tool, but a subject that can resonate with value, "I like Zhihu's page very much, with a faint light blue and warm text, 'If the unknown is the source of fear, then curiosity is a powerful weapon against fear'." 'Empathize with the world, so that you can always maintain a soft and strong heart', I am willing to stay on this platform more with such words".
Both technological media and human subjects are striving for the "perpetuation" of the media network that is jointly formed. If there is no human memory, AI memory cannot be born, and it is precisely because people can adjust the deviation of AI memory that the algorithm can continuously optimize the calculation program. Artificial intelligence must start from the memory needs of human beings, so that AI-based memories can be more smoothly integrated with individual memories, and the rejection of human integration of the two memories can be reduced as much as possible. Similarly, AI has strong autonomy in this media network, such as the right to choose memory glimmers and assign meaning to them, the right to repair memory biases, and the right to connect parallel memory worlds. In the face of the construction of individual memories by artificial intelligence, human subjects are not completely insensitive, let alone hand over the key to the memory palace.
6. The main body of AI-based memory escapes
(1) AI is me: reflect on the subjective illusion
Compared with traditional media memory, AI-based memory is more hidden from the construction of human subjectivity, and users often cannot perceive its influence in time, but unconsciously internalize it into their own memory, thus producing the illusion that "AI is me". On the one hand, this is due to the fact that the personal data processed by AI memory is often actively produced by users, and on the other hand, artificial intelligence often has both objective data presentation and warm text expression when integrating memory, such as "You played this song on a loop 58 times on November 14, and you remember this song no matter what happened that day". According to the diary keepers in this study, AI-based memories are "vivid and interesting", no different from the content of personal memories.
In the annual report of station B, it was said that I watched the National College Debate League 8 times this year, which triggered my memories of playing the debate competition last month, and now it seems that it was a hard but warm and beautiful time. I took the initiative to search and view these data, I think these data that summarize my "B station career" this year are very interesting, and I will also search for more videos I like from the direction of my preferences that it launched. (Recorder 55, 19:03 on December 31, 2021)
Technology is often not easily detected when it is used smoothly by people, and it is only when the subject feels uncomfortable that attention is paid to the alienity of the technology. When the subjects of our study change from "insensitive" to "sensible" when the intrusion of AI-based memory is "infelt", we will reflect on the illusion that "AI is me". This reflection originates from the interruption of man's "hands-on state" of technology, and the materiality of technology is revealed, so that the subject's perception is disturbed. From the diary of Recorder 38, we can find that the extraction of AI-based memories is not as natural as mobilizing individual memories, "When I want to view my personal footprints (in the AutoNavi map), I have to log in to view the specific footprints, which I find very annoying and unnecessary." The unnaturalness of this kind of mobilization prompted the subject to begin to recognize the "transaction scam" of the technology, "I am a little worried about my data privacy, because everything I listen to can be seen by NetEase Cloud", and the user's awareness of the boundary of their own memory reflects their rejection of AI-based memory chimerism.
(2) Verbal Escape: Reconstruct the Memory Constellation
The algorithm transforms the social communication behaviors such as sharing, creativity, and empathy of users using the online platform into the basic means of production in the production mode of the platform economy, and echoes the quantitative characteristics of the capital world in the form of data flow. The algorithm divides users into different types through quantitative calculation, and when facing the same group, the AI-based memory outputs standardized and customized copywriting in the expression form. AI-based memory connects the automatically generated warm copywriting with individual memory, connects individual memory to the social macro background, and sublimates the social value of the platform, which is essentially to stimulate the reproduction of users and create more traffic that can be converted into commercial capital. From the diary materials in this study, it can be seen that AI-based memory has caused dissatisfaction among recorders because of sensationalism and attention, "I think Alipay's bill calculation deliberately retains an incomplete algorithm and creates billing anxiety to win attention for itself, does it really clearly show the benefits?" In this regard, users will take ways to question and criticize, reduce consumption, terminate services, etc., to get rid of the words of AI-based memory, and resist the embedding of technological consciousness in individual memory. Recorder 4 recognized the commercial inducement and criticized, "In 365 days a year, (station B's 2021 annual report shows) I bought a big membership for 191 days, and I spent a total of 95 yuan on the conversion, I question station B wants to guide me to rush for another annual big member"; Recorder 78 directly reduces consumption to counter the power of capital, "the significance of Meituan's annual bill is to reflect on and summarize its own consumption habits and consumption structure"; Recorder 38 rebelled in a more radical way, "[in the Taobao bill] 'my Taobao personality' made me angry, and it was summed up as 'decisive and resolute', which I felt was capitalist rhetoric, and I didn't even want to shop on Taobao anymore."
After getting rid of the memory language, the user begins to reconstruct the memory constellations that are similar and structurally different, as described by Walter Benjamin, each star is independent but unified in the whole constellation. On the one hand, the stars in the Memory Constellation are identical as a whole. AI will push similar memories to the same type of users, and some users will recognize this type of memory, "(QQ annual social image) in the 'heartbreaking is you, and it is you who makes people love' This comment looks very unique at first glance, until I saw other friends in QQ space who also commented like this, I realized that I thought too much, but it is just the product of big data algorithms, and people with distinctive personalities are just a string of similar data and code in the eyes of big data." On the other hand, the stars in the memory constellation are all personalized in their specific morphology. Although AI memory seems to use the same text to apply different data of individuals, behind each piece of data represents the different experiences and experiences of each user. On the basis of the typed memory output by AI, users combine the specific situation and emotional cognition at that time to construct differentiated memories and create a unique and vivid memory constellation. Recorder 49 explains the uniqueness of the memory constellation very well, "(Weibo annual report shows) 'In July 2021, you sent 159 Weibo posts in one go, and your writing made Li Bai accept three points. This summary looks hilarious and witty, but in the context of the time, you will feel that AI is stupid, angry, and does not understand basic human emotions. My hometown is Zhengzhou, Henan Province, and I experienced heavy rain in July, and reposting Weibo is basically helping others for help."
(3) Emotional alienation: create a memory seal
After the integration of AI memory and personal memory, it has a stronger ability to mobilize emotions. AI-based memory customization presents personal memories in a certain period of time in the past, with a sense of ritual and surprise, which stimulates the user's emotional positive energy. Recorder 11 used "moving" to summarize the emotions evoked by AI-based memories, "I listened to "Dancing with your ghost" 104 times a day at the last moment, which reminded me of the torment of writing research reports in the summer, the days when I listened to songs for fun, and when I saw QQ's annual song listening report, I was still quite moved." Successful emotional arousal is often characterized by the act of sharing, for example, the recorder 14 was emotionally infected, "will want to share (the QQ Music Annual Report) with his classmates and friends, and to some extent will also resonate". This kind of sharing aims to achieve self-identity and group identity, and similar individual emotional experiences gather into a collective force to form an emotional community, which stimulates the continuous burst of emotional energy. The construction of group symbols is the result of emotional mobilization, such as the "personal image business card" mentioned by the recorder 4 has become a familiar symbol for the majority of music lovers, "NetEase Cloud Music Annual Report by analyzing our private data, for us to speculate on the psychological age, summarize the key words of listening to songs, and make personal image business cards, and its private customized charm has boosted this national swiping screen." These symbols can mobilize the emotional memory of the individual and quickly coalesce into collective action and release emotional energy the next time they encounter a similar situation.
The emotional mobilization process of AI-based memory is not all smooth sailing, and there are always retrograde people among our research subjects, sealing the door of the memory world and escaping from emotional mobilization. Freezing the memory entrance is the first step for users to create memory seals, users are unwilling to view AI memories and remain emotionally alienated from them, so they cannot produce effective emotional energy, "I have not taken the initiative to view these data, it was officially pushed to me by Douyin, and the data of 'One Week Achievement' are not helpful to my study and life." Recorder 19 also made a sarcasm on the spread of AI-based memories, "NetEase Cloud Music has to conduct various summary tests two or three times a year, and for teenagers who are keen to show themselves, its transmissibility is no less than that of the flu virus." If freezing the memory entrance is the user's resistance to technological shaping, then isolating the technical consciousness is the subject's more resolute rebellion against AI-based memory. "QQ (2021 Social Image of the Year) awarded me the title of 'Business Expert', but I didn't want to get it at all, social software wasted too much of my time. One of my big New Year's wishes is to use this stuff less." Users not only cut off the chimera of individual memories by dissolving the symbols of AI-based memory, but also reshape memory tags with the initiative of the subject's consciousness to challenge the construction logic of AI-based memory, "I don't like the labels that the Zhihu annual report puts on me according to the topics I often browse, most of the time I keep reading these answers because I don't understand but am a little interested, not because I belong to the label image it says, which is problematic in planning, My image should be analyzed according to the type of answer I make".
From the reflection of memory subjects, the reconstruction of memory frameworks to the isolation of memory channels, human subjects adopt a variety of thinking plans and action strategies when they encounter AI-based memory. Similar to the previous changes of technological media, the cooperation, negotiation and confrontation between human subjects and technological media continue in the construction of AI-based memory. The difference is that human beings have never faced such a realistic "technological incarnation" before, and this encounter is destined to be full of mutual attraction and entanglement between human subjects and technical subjects. Because of this, it is more difficult for humans to keep a distance from AI-based memory and reflect on subjectivity compared to traditional media memory, and when AI-based memory fully penetrates into daily media use and continues to extend from individual memory to circle memory through emotional networks, technical disconnection with it often means deep social isolation. In this sense, the various manifestations of the escape of the above subjects have some instinctive and temporary meanings.
VII. Conclusion
The AI-based memory investigated in this study uses individual network data as the material and artificial intelligence algorithm as the construction rule, and comprehensively penetrates the situational memory, program memory, and working memory of human subjects. By lighting up the glimmer of memory, AI-based memory breaks away from the generative framework of grand narratives; Through the repair of memory bias, individuals are given more agency; Through the construction of parallel memories, individuals can move freely between the real world and the virtual world. All this seems to lead to a media memory space full of differences and individualization, thus completing the subjective chimerism of technological media on people without a trace. If this chimerism can achieve a "perfect match" with the embodied memory of the individual, then there will be no escape behavior of the memory subject due to the imbalance of perception in our study. From this point of view, the game between individual memory and media memory, and the interaction between human subjects and technological media are still being staged in the era of artificial intelligence, but this time the repertoire is different.
As mentioned above, traditional media memory realizes the structuring and institutionalization of individual memory through the use and construction of cultural scripts, and to a certain extent, individual memory exists in the form of collective memory and social memory in traditional media memory. As a new memory agent, the biggest difference between artificial intelligence and previous media is that it uses algorithm programs instead of cultural scripts to quantify and communicate fragmented network data. The so-called contract, which is to express or measure the characteristics that are usually represented by different units according to a common metric standard, artificial intelligence may be the most refined and perfect communication tool after money. In the process of abstracting and restoring individual network data, the connection between appearance and the empirical world is obscured, uncertainty is absorbed, and AI-based memory establishes a new system for integrating individual memory. In this system, artificial intelligence and its algorithms give uniformity to otherwise differentiated things, but this unity is often considered to be a property of an object rather than a product of quantification, and algorithmic programs are also thought to be only objectively calculating or measuring something rather than generalizing different entities. As a commensurable thing, AI-based memory is more accurate and standard than traditional media memory, and it is also more in line with the needs of modern society and commercial capital for comparability and circulation, so that individual memory becomes a "customizable resource", and a new model of the relationship between technological media and human beings will also emerge.
As Kittler argues, the more intelligent the media develops, the more degraded human subjectivity becomes. In the period of natural media, people directly feel the external knowledge and experience through the senses, construct self-identity and basic cognition of external objects through the media, and establish the subjective status of human beings in the world. After the advent of machine media, the media can reproduce the real-time dynamics of human beings, and synchronously generate digital memories that track individual footprints, and people's subjectivity in the real world can be extended to the virtual world, and each cluster of media memories corresponds to a specific cultural schema. In the era of intelligent media, artificial intelligence with the ability of independent learning and evolution breaks the boundary between man and machine, embeds individual memory, and reproduces it in a new form, creating AI memory of human-machine symbiosis, from which human cyber subjectivity is born. Cyber subjectivity is the deep evolution of human subjectivity by artificial intelligence, and it is an advanced form of human-machine interdependence. From the emotionally conscious replicant in Blade Runner to the android made of real data in Black Mirror to the human-machine humanoid in Alita: Battle Angel, the idea of cyber subjectivity is endless. The latest progress of artificial intelligence is turning the above assumptions into reality, and the future of technology is approaching step by step, and it can be expected that the cyber subject will reconstruct the role of people as memory subjects in three aspects: the cyber subject may break the physical limitations of memory containers, so that people and intelligent technology carriers can run in to form a new body structure, and extend the functions of various organs to store and display memory information; By establishing a new memory system for the permanent and complete preservation of memory content, the cyber subject not only gathers the memory glimmer into the memory strong light, but also realizes the transformation to the constant memory light, and eliminates people's fear of memory forgetting. Cyber subjects may even directly share visual, auditory, olfactory and other memory content, eliminating bias in memory sharing and completing lossless memory exchange.
Through the construction of quantitative convention and cyber subjectivity, artificial intelligence gradually constructs a memory world with standard structure, clear details and stable connotation. The irony may come from this, where early humans created artificial intelligence according to their own intelligence standards, and when this imitation of human intelligence develops to an uncontrollable level, it will cause a "technological backlash" to human beings, breaking the sole subjective position of human beings in the world. As the spokesperson of algorithms, AI-based memory connects scattered resources into a more stable network, and then becomes a "mandatory passing point" on the actor's network. At this point, people, things, and media are in a kind of relationship network, and they are connected to each other for the stable operation of the media network system. In this multi-agent world, people, like other things, will be demasked and centralized, and their respective values will be transformed through the processes of development, distribution, and transformation, and become objects in the eyes of other subjects.
In the face of the menacing technological accumulation, the value of this research ultimately raises the question of how human beings can anchor themselves in the world of floating subjects in the future. In this regard, the media memory retained by human beings as the only subject may help human beings shuttle through the network of technology, after all, these irreducible memories have been used as a concrete manifestation of the subject's consciousness, and have also shaped the value orientation of technological media. If the golden age of subjectivity is coming to an end, then the memory it retains as a new element to reshape the multi-agent relationship can not only help humanity resist "technological oblivion", but also may reactivate the gene of human subjectivity to form a unique and vivid node in the human-machine network. In this sense, the memory formed based on various specific factors such as culture and ethnicity is particularly precious, and the construction of cultural memory and collective memory by intelligent media can also become an important direction for future research.
This paper is the research result of the general project of the National Social Science Foundation of China, "Research on the Generation Mechanism and Governance Strategy of Network Social Mentality from the Perspective of Emotional Communication" (Project No.: 23BXW025).
Originally published in Journalism and Communication Research, Issue 1, 2024
Due to space limitations, the official account will be left with annotations, and the full version can be found in the publication