By 'augmenting human intellect' we mean increasing the capability of a man to approach a complex problem or situation, to gain comprehension to suit his particular needs, and to derive solutions to problems." So wrote Douglas Engelbart in 1962, initiating a narrative of augmentation that runs through hypertext's history. To augment is to amplify something: our ideas, our activities, our thoughts. This paper explores hypertext's relationship with computers as intentional machines, its relationship with artificial and human intelligence, and what role it can play in negotiating between the two.

• Hypertext; • Artificial intelligence; • Creative intelligence;

Robotics, AI Ethics, Hypertext, Cultural Impact

Dr Sam Brooker. 2024. Computer, Enhance!: Augmentation, Ideation, Hypertext. In 35th ACM Conference on Hypertext and Social Media (HT '24), September 10–13, 2024, Poznan, Poland. ACM, New York, NY, USA, 4 pages. https://doi.org/10.1145/3648188.3678216

The parallel developments of artificial intelligence (AI) and hypertext offer contrasting approaches to the augmentation of human intellect and the evolution of our relationship with technology. While AI has often been viewed (at least in popular fiction) primarily from the perspective of autonomous decision-making and its potential to either liberate or dominate humanity, hypertext offers a contrasting approach that emphasises connection, interpretation, and the expansion of thought processes. This paper explores the historical and cultural contexts that have shaped our understanding of AI and hypertext, examining how these technologies can complement each other in augmenting human thought.

By exploring the philosophical aspirations of hypertext as a method of inquiry and its potential to demystify AI, this paper aims to highlight the importance of human-centric design in the development of intelligent systems. This approach not only addresses the existential anxieties surrounding AI but also reaffirms the role of hypertext in fostering a more transparent and collaborative interaction with technology.

Our science fiction – our cultural imagination - is replete with mechanical servants. Popularised by Czech writer Karel Čapek in his 1920 play RUR, our common term robot has Slavic roots, sometimes neutrally translated as being associated just with labour but more accurately traced through the idea of forced work and drudgery, even ultimately (in Old Slavic) to slave. Referring to himself in the third person, Čapek would later decry efforts to equate physical robots with the more philosophical robots he intended in his original work: "With outright horror, he [Čapek] refuses any responsibility for the thought that machines could take the place of people, or that anything like life, love, or rebellion could ever awaken in their cogwheels" [1]. In the same year Čapek wrote this article, Nikola Tesla wrote in similarly agnostic terms about putative mechanical servants: "In the twenty-first century the robot will take the place which slave labor occupied in ancient civilization. There is no reason at all why most of this should not come to pass in less than a century, freeing mankind to pursue its higher aspirations" [2]. Originally derived from the Japanese expression 3K (describing work often associated with migrant workers or those belonging to a traditionally feudal underclass), the widely used neologism 3D/4D is used within the robotics industry to describe work (Dirty, Dangerous, Demanding, Dull, Demeaning, alongside various others) that can be more humanely undertaken by machines (though the loss of this work has its own significant moral consequence, as we will see below). We need not invoke Asimov's laws of robotics to make the simple point here that such machines are understood as agents whose primary role is to serve us as we choose.

When they are given a voice, this coding (in the cultural studies sense) tends to continue. When Captain Kirk – a problematic 1960s icon of virile, take-charge masculinity – makes demands of his computer, it is with a secretarially coded tone that the obedient computers of Star Trek respond. Writing of Apple's Siri voice assistant, Liz W. Faber notes that "her voice is female coded. . . because Siri's gender—as well as the gender of the fictional computers before her—is an expression of deeply rooted, complex, problematic, and problematizing gender roles" [3, p. 178].

Here the computer (as with other machines) is reactive, not proactive (it is worth noting that when computers takes charge, the voice is often male - HAL 9000, SkyNet, even Auto from the Pixar movie Wall-E – though there are exceptions, as in the film Ex Machina.) Our digital assistants are dormant until we give them a purpose: we use wake words to issue commands, after which the machine recedes once more to await instruction.

Intention and purpose are the domain of human agents, action the domain of the machine. Where machines go wrong, it is often by distorting or misrepresenting our intentions: the aforementioned HAL 9000 refusing to open the pod bay doors in 1968 film 2001: A Space Odyssey; VIKI in 2004's I, Robot; the pragmatic computer of Harlan Ellison's 1967 I Have No Mouth And I Must Scream. In each case it is the distorted mirroring of our own ethical constructs that provides the threat, an obedience so profound as to become destructive.

The alternative, of course, is a ruthless AI that sees us fundamentally as a threat or an inferior species, as seen in the Terminator and Westworld franchises among numerous others. In the real world, optimists and pessimists have proposed this scenario as a possible future: "Will robots inherit the earth?" wrote Marvin Minsky [4]. "Yes, but they will be our children. We owe our minds to the deaths and lives of all the creatures that were ever engaged in the struggle called Evolution." Offering a slightly more optimistic view than handing the torch to our digital antecedents, Norbert Weiner [5] argues that just as the first industrial revolution "devalue[ed] the human arm by the competition of machinery" so the computer revolution "would devalue the human brain". If you were wondering where the optimism comes in, it's here: "just as the skilled carpenter, the skilled mechanic, the skilled dressmaker have in some degree survived the first industrial revolution, so the skilled scientist and the skilled administrator may survive the second." Cold comfort, indeed.

In the past we have been both terrified of machines with agency, and highly skeptical: machines whose logic convinced them to sacrifice humanity for its own good can usually be thwarted by exploiting that same logical mindset, as the aforementioned Captain Kirk does in four separate episodes of Star Trek. Those ruthless artificial intelligences that sought to engineer our destruction are in turn often overcome by our tenacious, adaptable human spirit.

"For the marchers of the Free Speech Movement," wrote Fred Turner [6], "as for many other Americans throughout the 1960s, computers loomed as technologies of dehumanization, of centralized bureaucracy and the rationalization of social life, and, ultimately, of the Vietnam War. Yet, in the 1990s, the same machines that had served as the defining devices of cold war technocracy emerged as the symbols of its transformation." Paranoia about the intervention of computers into our lives around the time Ted Nelson first coined the term hypertext had been replaced by optimism in precisely the decade that hypertext began to find general use. Perhaps coincidentally this is the inverse of the two periods referred to as the AI winter (1974-1980 and 1987-2000) when funding and support for AI development dried up; that said, Paul N Edwards [7], in his history of cold war computing, explicitly identifies renewed interest in AI as emerging from increased funding for cold war defence technology under the Strategic Defence Initiative. Hypertext's best decades were the same decades technology as a whole was being celebrated, while AI tended to arise during periods of comparatively disinterest and paranoia.

Paranoia is back, however, and negotiating this anxiety about intentional machines has been thrown into sharp relief since OpenAI offered ChatGPT to an astonished public. "Public debate about risks related to Artificial Intelligence," write Turchin et al. [8] "is intense, but the discussion tends to cluster around two poles: either mild AI risks like AI-driven unemployment, or extinction risks." Any new technology in which even mild risks represent potential loss of employment is bound to yield general anxiety, while the idea of existential risk (however strongly derived from our popular fictions) is apt to make even the most evangelical technophile nervous. As a source of technostress - "a negative psychological response to the use (and abuse) of technologies, as well as the harmful effects of the implementation of technologies within the workplace" – AI at the present moment is hard to beat.

Both materially and in nomenclature, the idea of AI as an autonomous decision-making entity has been downplayed by principle actors in this space: Microsoft recently launched its AI assistant under the title Copilot, foregrounding both its intelligence but (crucially) its subservience to the user; the ubiquitous chatbox interface used by most LLMs affords a feeling of control over interaction, a sense that we are engaging in real conversation. When users push for restricted or otherwise inappropriate content, ChatGPT's tone is obsequious and apologetic.

Nonetheless there is something unsettling about the idea that computers could have intentions or steer conversations – more so, given how reluctant providers of such services are to acknowledge this level of control. The small and unobtrusive addition of the disclaimer ChatGPT can make mistakes. Check important info. below the chatbox feels like an afterthought, equivalent in prominence and importance to a begrudging cookie notification. The sense still exists that while Chat GPT will respond, they may even do all the work – we still issue the command. The idea of breaking this fundamental rule feels wrong, since ideation and insight is our (the user's) territory. "If AI were ever to succeed in creating a whole personality inside the computer," wrote Jay Bolter [9], "the human user would become a mere passive reader who listened to the monologue provided by the machine." Agency and the rest resided – for Bolter, at least – firmly within hypertext.

In their 2019 paper Hypertext as Method, Atzenbeck et al. outline a range of ways in which hypertext can be seen as a method of inquiry, rather than a type of system [10]. They locate as the centre of current work in hypertext the idea of augmentation, going on to cite Douglas Engelbart in saying: "By 'augmenting human intellect' we mean increasing the capability of a man to approach a complex problem situation, to gain comprehension to suit his particular needs, and to derive solutions to problems" [11]. In the aforementioned paper Minsky, too, talks about "augmenting our brain" both mechanically but also metaphorically.

For the purposes of this paper, we may consider augmentation to be the enhancement of digital text with interactive elements (linking, annotation, multimedia integration), allowing users to engage with and navigate content in a non-linear manner. By connecting disparate pieces of information, users enjoy a more dynamic working environment, facilitating deeper understanding and more effective information retrieval.

This idea is widely discussed within the hypertext community. In keeping with the cybernetics of his tutor and sponsor Joseph Licklider, Engelbart's report speaks of "developing means to augment the human intellect . . . [and] help man apply his native sensory, mental, and motor capabilities". He goes on to speak of computers as "clerks", helpers whose job is to assist the user. This characterisation of the benign computer recalls Vannevar Bush's army of low-level machines from his influential 1945 article As We May Think and various works by Ted Nelson that speak of "versatile gizmos" of one stripe or another working to support the processes of ideation and development [12]. It is present in Landow's idea [13] that "hypertext embodies the very structure of our thinking process, allowing us to explore ideas through a network of interconnected information" or Bolter's argument [9] that "hypertext allows us to externalize our thoughts and link them together, expanding the boundaries of what we can know and understand." Visual design expert Manuel Lima [14] notes that "in the age of hypertext, our ability to connect diverse pieces of information through networks of links enables a deeper, more comprehensive understanding of complex subjects," as Howard Rheingold [15] once assured us that "hypertext fundamentally augments our cognitive capabilities by allowing us to navigate and synthesize vast amounts of information in ways that linear text cannot."

It is the view of this paper that beyond even its application as method of enquiry, hypertext's philosophical aspiration toward the augmentation of human intellect – present at every level of its activity – is one way in which our community can contribute to the development of what is colloquially described as artificial intelligence. Hypertext as a tool for augmentation is present from the start in Nelson's Project Xanadu, envisioned as a comprehensive hypertext system aiming for interconnected, editable documents, and Douglas Engelbart's NLS (oN-Line System), which introduced hypertext editing and linking within a collaborative environment. It can be found in Apple's HyperCard in the late 1980s, enabling users to create interactive stacks of virtual and linked 'cards' - and of course in the World Wide Web, which revolutionized global information sharing and where the bulk of knowledge exchange occurs today.

Hypertext as method is about extension, about expanding the capabilities of text; in turn, expanding the possibilities it offers. Mark Andersen and David Millard [16] note as one axis of hypertext scholarship Manual vs. Automatic, which "attempts to address the question of whether hypertext is augmenting our thinking or doing our thinking for us, which echoes to some extent the distinction between Tools for Thought and Knowledge Representation." They note that this question "seems equally as relevant today, especially with recent advances in AI." Perhaps for a wider public increasingly exposed to the potential perils of the thinking being "done for them" hypertext can move in the other direction (a shift already reflected in the increasing presence of book history at ACM Hypertext conferences.)

Both the artificial intelligence and hypertext communities have made claims regarding their similarity to the patterns of thought. Where pioneering AI scholars like Paul Churchland [17] suggested that neural networks might function like a brain, however, hypertext scholars tended to focus on the way in which their systems helped visualize or represent the patterns of thought. Here lies a significant distinction: between the mimicry of thought and its representation. We may recall the double logic of remediation, Bolter and Grusin's idea that our culture "wants both to multiply its media and to erase all traces of mediation" [18]. LLM-oriented AI in its current state erases the process of thought, revealing only the outcome; hypertext expands and reveals that process, showing the many paths that can lead from a single statement.

"Any nitwit can understand computers," continued Ted Nelson [12], "and many do. Unfortunately, due to ridiculous historical circumstances, computers have been made a mystery to most of the world." Hypertext is fundamentally about connection, interpretation – it is an architecture for discovery and exploration. Our current AI tools are often opaque and frustrating, oriented around outputs and obtuse about process. Hypertext has the opportunity to provide a more human, more kinaesthetic and visual vector for users. It can – as it does in spatial recommender system SPORE – provide an interface for the engineering of prompts, for example. As Roßner et al. [19] write in a paper discussing SPORE: "Text generation offers the author automation, generating output based on input parameters." The alternative (afforded by spatial hypertext and an in-built recommender system and knowledge graph) offers an augmentation of the creative process.

The challenge, of course, is in what way and how to communicate the importance of hypertext for artificial intelligence communities. Engagement with adjacent communities working in UX is one route, reaffirming aspects of hypertext oriented around the organisation and presentation of information. Reconnecting with creative communities like the Electronic Literature Organisation and other venues for digital poetics would also affirm the place of creative intention, as would continuing to develop strong relationships with book history and those working in scholarly editions. Earlier work on hypertext and pedagogy – as a way to structure and present knowledge – might also see renewed interest, as we seek new ways to challenge students and interrogate work that go beyond the traditional linear essay. General purpose AI has won the war of acceptance; hypertext can help it move from threatening cultural necessity to readable, explicable assistant.

"Tools for argumentation," write Anderson and Millard [16], "were Hypertext's response to the second Summer of AI, a quest for clarity in the face of bias, and a way of encouraging users to engage in more structured thinking." While claims around AI bias are complex and multifaceted, the encouragement for hypertext scholarship to engage in more structured thought and centre the human in our endeavour has as much currency as ever.

K. Čapek, "THE AUTHOR OF THE ROBOTS DEFENDS HIMSELF," LIDOVÉ NOVINY, 9 June 1935.

N. Tesla, "A machine to end war," Liberty, pp. 5-7, 9 February 1935.

L. Faber, The Computer's Voice, MN: University of Minnesota Press, 2020.

M. Minsky, "Will Robots Inherit the Earth?," Scientific American, pp. 109-113, October 1994.

N. Weiner, Cybernetics: or Control and Communication in the Animal and the Machine, Cambridge, MA: MIT Press, 1948.

F. Turner, From Counterculture to Cyberculture, Chicago: University of Chicago Press, 2006.

P. N. Edwards, The Closed World: Computers and the Politics of Discourse in Cold War America, Cambridge, MA: MIT Press, 1997.

A. Turchin and D. Denkenberge, "Classification of Global Catastrophic Risks Connected with Artifical Intelligence," AI and Society, vol. 35, no. 1, pp. 147-163, 2020.

J. Bolter, Writing Space: Computers, Hypertext, and the Remediation of Print, 2nd ed., Mahwah: Lawrence Erlbaum Associates, 2001.

C. Atzenbeck and P. Nurnberg, "Hypertext as Method," New York, NY, 2019.

D. Engelbart, "Augmenting Human Intellect: A Conceptual Framework," October 1962. [Online]. Available: https://www.dougengelbart.org/pubs/augment-3906.html. [Accessed 14 June 2024].

T. Nelson, Computer Lib/Dream Machine, 2nd ed., Redmond, Washington: Tempus Books, 1987.

G. Landow, Hypertext 3.0: Critical Theory and New Media in an Era of Globalization, 3rd ed., Baltimore, MD: Johns Hopkins University Press, 2006.

M. Lima, The Book of Trees: Visualizing Branches of Knowledge, Princeton, NJ: Princeton Architectural Press, 2014.

H. Rheingold, Tools for Thought: The History and Future of Mind-Expanding Technology, 2nd ed., Cambridge, MA: MIT Press, 2000.

M. Anderson and D. Millard, "Seven Hypertexts," New York, NY, 2023.

P. Churchland, "Cognitive activity in artificial neural networks," in An Invitation to Cognitive Science, D. N. O. &. E. E. Smith, Ed., Baltimore, MD, MIT Press, 1990, pp. 3-372.

J. Bolter and R. Grusin, Remediation: Understanding New Media, Baltimore, MD: MIT Press, 2000.

D. Roßner, C. Atzenbeck and S. Brooker, "SPORE: A Storybreaking Machine," New York, NY, 2023.