Authors:
Chameera De Silva, Thilina Halloluwa
Artificial intelligence has long influenced how humans interact with computers. Significant applications that have augmented human capability and intelligence include search engines, virtual assistants, and personalized recommendation systems [1]. In this context, augmenting means enhancing human abilities and experiences through AI—complementing and amplifying human potential without replacing human creativity or connection. A smartphone, for instance, might appear almost wizard-like to our ancestors, granting access to boundless information and enabling feats they could scarcely imagine, such as retrieving the precise distance to the sun and allowing us get dinner delivered to our doorstep with a few simple taps. In 1968, Douglas Engelbart saw the potential for computers to act as tools that augment human intellect, essentially jump-starting how we think about our interaction with computers. He famously stated that "the better we get at getting better, the faster we will get better," and that statement seems truer than ever with the rate of AI advances and the use of AI in advancing AI. Large language models (LLMs) represent a transformative leap within AI, introducing novel interaction paradigms between humans and computers. They offer a new, unparalleled way to automate processes, access boundless amounts of information, and cut tiny menial text-based tasks out of people's lives.
Insights
→ Large language models are transforming HCI by enhancing creativity, efficiency, and accessibility.
→ Overuse or inappropriate application of LLMs can lead to social isolation and diminished human connection.
→ The focus should remain on using LLMs to augment human potential while addressing ethical and practical challenges.
A notable example is ChatGPT, a widely adopted application that has fundamentally changed how users interact with technology. An example of its revolutionary impact was observed when, following its release, a significant decline in visits to platforms like Stack Overflow occurred (Figure 1). It underscored the substantial impact LLMs have had on the broader digital landscape [2]. The advent of LLMs, such as GPT-3.5 Turbo and GPT-4, and their availability via OpenAI's API, have unlocked unprecedented opportunities to enhance user experiences. This article examines the intersection of human-computer interaction and LLMs, analyzing how these models are revolutionizing user interfaces, improving accessibility, and presenting new challenges and opportunities in the design and interaction domains.
 | Figure 1. Stack Overflow's traffic over time. |
The Use of Text for Manipulation and Creation
LLMs have revolutionized natural language processing, enabling advanced text generation and manipulation. These models are now widely used across various domains, including content creation, data synthesis, and automated reporting. Their ability to produce humanlike text has expanded applications in automated journalism, creative writing, summarization, translation, and code generation, boosting productivity and creativity in technical and nontechnical fields.
Companies are increasingly using LLMs to automate media production, such as writing scripts, composing songs, and generating Web content. By streamlining content creation, these models reduce costs, scale output, and maintain consistency in style and voice, making them invaluable in industries like entertainment and digital marketing.
LLM Agents and Autonomous Task Execution
The development of LLMs has given rise to a new class of autonomous agents that can perform a variety of tasks on behalf of users. These LLM-powered agents are designed to understand complex instructions and execute them with minimal human intervention, making them critical tools in areas such as customer support, virtual assistance, and workflow automation [3]. For example, LLM agents can be deployed in customer service environments to handle routine inquiries, resolve issues, and provide personalized recommendations, thereby reducing employees' workload and improving response times. Additionally, these agents are increasingly being used in personal and professional contexts to automate tasks such as scheduling, email management, and data analysis. By leveraging the capabilities of LLMs, these agents can operate with a level of autonomy that was previously unattainable, allowing for more efficient and effective task management. The reliance on LLM agents, however, also raises important questions about the delegation of decision-making processes to AI systems and the implications for human oversight and accountability.
Multimodal LLMS and Accessibility
The integration of multimodal capabilities into LLMs has significantly enhanced their potential to improve accessibility across various domains. Multimodal LLMs are designed to process and generate information across different types of inputs, such as text, images, and audio, enabling more comprehensive and contextually aware outputs. One key application of this technology is in the conversion of natural language into structured query languages (NL to SQL), which allows users to interact with databases without requiring advanced technical knowledge. This capability democratizes access to data and empowers users to perform complex queries by using simple text commands. Another important application is in the realm of vision assistance, where multimodal LLMs can analyze information and provide descriptive text or auditory feedback to visually impaired users. This can significantly enhance the independence and quality of life for individuals with disabilities, allowing them to engage more fully with digital and physical environments. The convergence of multimodal inputs within LLMs opens new avenues for creating more inclusive and accessible technologies, bridging the gap between human abilities and machine capabilities.
Potential Downfalls: The Overextension of LLMS
While the capabilities of LLMs are vast, their overextension into domains where they may not be well suited can lead to significant challenges. The widespread enthusiasm for LLMs has led to their deployment in areas where their utility is questionable and where they may introduce unforeseen complications. Businesses, eager to capitalize on the AI trend, have at times implemented LLMs in ways that do not align with the specific needs of their operations or user bases. Examples such as the Rabbit R1 project [4] and Humane Ai Pin [5] demonstrate the pitfalls of integrating LLMs without a clear understanding of their limitations and the contexts in which they operate. In these cases, the push to incorporate AI-driven solutions led to outcomes that were either ineffective or counterproductive, highlighting the importance of critical evaluation of LLM applications and their interactions with humans. Moreover, the growth of LLMs in contexts where they may not be necessary or appropriate can lead to user frustration, decreased efficiency, and a dilution of the perceived value of these technologies.
Decreased Human Interaction and the Risk of Isolation
The rise of automation through technologies like LLMs is transforming how we engage with the world, streamlining services and making interactions more efficient. However, this growing reliance on automation has sparked concerns about its impact on human connection, raising the possibility of increased social isolation and a diminished capacity for interpersonal communication.
In industries such as retail and hospitality, automation has become a double-edged sword. Wendy's AI-powered drive-through system is one such example [6]. While the fast-food chain revolutionized the consumer experience by offering faster and more accurate order processing, it also eliminated opportunities for the casual, often meaningful interactions that occur in human-centered transactions. This trend reflects a broader societal shift where the convenience of machine-driven efficiency increasingly takes precedence over the richness of human engagement. The result is a world where even mundane interactions, like ordering a meal, become transactional, further distancing us from one another.
 | The Humane Ai Pin. |
Throughout history, humans have often prioritized convenience over quality. We chose microwaves over stoves, fast food over home-cooked meals, and single-use plastic bags over reusable totes. LLMs now present a similar choice, offering unprecedented convenience and capability but at the potential cost of genuine human connection. These systems are capable of performing tasks requiring basic reasoning and, in doing so, are poised to replace countless human roles in customer service, education, and beyond.
As LLMs become more embedded in daily life, the risk of isolation grows. Interactions with AI are increasingly preferred due to their comfort and efficiency, particularly as social anxiety rises. For many, the nonjudgmental nature of an LLM agent can seem more appealing than the unpredictability of a conversation with a person. Yet there is an inherent danger in normalizing artificial interactions over human ones. This growing reliance on AI could shift societal norms, favoring efficiency over empathy and isolation over community.
The long-term consequences of this shift could be profound. As face-to-face interactions diminish, we risk eroding the bonds that underpin our communities and personal relationships. Workflows and routines may become increasingly solitary, threatening the social fabric that makes us human.
To prevent this, we must strike a balance between embracing technological advancements and preserving the irreplaceable value of human connection. LLMs and similar technologies undoubtedly have the power to augment our capabilities, but they must be integrated in ways that complement rather than replace human interaction. By doing so, we can ensure that the march of progress does not come at the cost of the social bonds that sustain us, preserving humanity at the heart of our increasingly automated world.
Conclusion
The intersection of HCI and LLMs represents a moment of inflection in the ongoing journey to augment human potential. Echoing Engelbart's vision of augmenting the human intellect, LLMs have become powerful tools that amplify human creativity, efficiency, and accessibility. These systems not only revolutionize user experiences but also present opportunities to enhance individual capabilities in ways that were previously unattainable.
Realizing this potential, however, requires careful navigation of challenges such as trust, bias, and privacy. As we strive to create a future where technology and humanity converge seamlessly, the focus must remain on leveraging AI to enhance—not replace—human interaction, creativity, and intellect. In doing so, we ensure that the advancements in AI align with Engelbart's enduring vision of augmenting the human intellect for the betterment of humankind and avoiding the degeneration of human society.
References
1. Bender, E.M., Gebru, T., McMillan-Major, A., and Shmitchell, S. On the dangers of stochastic parrots: Can language models be too big? Proc. of the 2021 ACM Conference on Fairness, Accountability, and Transparency. ACM, 2021, 610–23.
2. Da Silva, L., Samhi, J., and Khomh, F. ChatGPT vs. LLaMA: Impact, reliability, and challenges in Stack Overflow discussions. arXiv:2402.08801, Feb. 13, 2024.
3. Huang, X. et al. Understanding the planning of LLM agents: A survey. arXiv:2402.02716, Feb. 5, 2024.
4. Pierce, D. The Rabbit R1 is an AI-powered gadget that can use your apps for you. The Verge. Jan. 9, 2024; https://www.theverge.com/2024/1/9/24030667/rabbit-r1-ai-action-model-price-release-date
5. Pierce, D. Humane officially launches the AI Pin, its OpenAI-powered wearable. The Verge. Nov. 9, 2023; https://www.theverge.com/2023/11/9/23953901/humane-ai-pin-launch-date-price-openai
6. Spessard, M. Leading drive-thru innovation with Wendy's FreshAI. Wendy's. Dec. 11, 2023; https://www.wendys.com/blog/drive-thru-innovation-wendys-freshai
Authors
Chameera De Silva is a clinical data scientist at Annalise.ai, a doctoral scholar, and an academic at James Cook University. With extensive experience in data engineering and clinical research, he specializes in designing, developing, testing, and deploying innovative applications. Proficient in writing code and algorithms, he brings expertise in building complex neural networks using various programming languages. [email protected]
Thilina Halloluwa is a lecturer in the School of Electrical Engineering and Computer Science at the University of Queensland in Australia. He has more than 15 years of academic and industry experience. [email protected]
This work is licensed under creative commons attribution international 4.0.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2025 ACM, Inc.
Post Comment
No Comments Found