1 Introduction

Human computer interaction (HCI), primarily, is the study of how people interact with computers. It has now grown into an independent research discipline with many researchers completely dedicating their research interests. HCI still remains at the core an interdisciplinary topic. Such an interdisciplinary nature has helped HCI to propagate into many independent dimensions such as UI/UX design, ubiquitous computing or visualization. Like other research disciplines, HCI intends to explore more areas of application.

One such potential application that HCI research intends to explore is the thriving artificial intelligence (AI) research. AI, broadly, is intelligence exhibited by machines. Artificial intelligence has found plenty of applications in research areas such as image recognition, speech processing, and image generation among others. These developments can be seen in self-driving cars, AI assistants and many other applications. These devices, applications and artefacts have found their way into everyday lives across multiple cultures. The ubiquitous applications of AI across digital devices have elicited the need of user interfaces. HCI-propelled technologies, such as UI/UX frameworks, which have been already successful with existing technologies, have the potential to prove very useful in these AI-driven technologies.

However, there remains a bridge for HCI to cross. A formidable critique of AI research tends to view AI as agents that have completely opposed interests with respect to humans. AI, in a more developed form, would compete with man for resources. In contrast, HCI historically has been viewed as an area of research that is dedicated to building computing systems that prioritize human interests. So, in a way, HCI stands directly opposed to AI. In this paper, we look at the relationship between AI and HCI. For this purpose, we briefly look at the history and analyse the discussion that has been taking place around HCI and AI integration. We also look at how AI has an impact on human lives, derive insights from them and conclude by suggesting a few implications of these insights into HCI research.

2 The critique of AI and the existing HCI–AI discussions

A principal part of the criticism of AI focuses on the fact that AI at some point in time will compete with, outlive humans and make them obsolete. This sort of argument is usually dismissed by AI researchers, other technologists and investors as some sort of witless petrified doomsday speculation. However, we think that there lies some sort of merit in the argument and countering this criticism can yield a lot of insights for the HCI community. Before diving in, it is imperative to examine some existing literature on HCI–AI integration. Winograd (4) writes that the AI and HCI communities have had opposing views of how humans and computers should interact. The work by Grudin (1) has described how both these fields received attention and funding alternatively across the 1950s to the 2000s. This apparently proved to be a very fertile period of understanding AI as well as for HCI. Shneiderman and Maes (3) have a very interesting discussion about the nature of the human–computer interface. Should the human interact with the computer as if it is another human or is it that it is not possible (with technical and philosophical constraints) to imagine a computer with human-like attributes? A major part of the debate was also between direct manipulation and interface agents. The basic idea is that whether agents learn the user's likes and dislikes and learn on behalf of them and whether the users really want that to happen. There are many interesting remarks in the discussion. Markoff (2) writes about the philosophical divide between John McCarthy and Douglas Engelbart. McCarthy set up his research system to come up with a superbrain kind of artificial intelligence (something that would be similar to, if not superior to) the human brain. Engelbart, on the other hand, thought of computing as a method to 'augment' the human. Such perspectives are really helpful in providing ground to the inevitable HCI–AI integration.

3 The AI critiques and the HCI–AI discussion

The basic arguments contained in most of the AI critiques boil down to a few ideas: What is to be done when machines take over redundant and low-level jobs? And what if machines and humans have opposing interests? We will try to analyse these basic ideas and see how the HCI–AI discussion pans out.

3.1 How humans have changed

Most of these discussions, if examined closely, maintain a very seemingly innocuous, but highly precarious position that humans throughout have remained the same, unchanged with time. This particular position serves as a substructure to most of the arguments governing the HCI and AI debates. To reassess this position, let us look at a saying in common parlance: 'First the man shapes the tool, and then the tool shapes the man'. Man's relationship with his tools is very intriguing. Say a man builds an axe and uses it to cut a tree. Now the axe has a prescribed technique to be used. It can only be used in such a way; it has to be hit hard on a tree in a direction and not just touched or rubbed on the surface of the tree to cut it. In other words, the axe has prescribed (or restricted) the degrees of freedom available to man to cut the tree, though allowing him to do so. Thus, the man, eventually, acquires the properties of the axe and the tool shapes the man. With this logic, technology as a tool can potentially change the human; take as an example, a technology like social media. Twitter takes up thinking with 140 characters at a time, Instagram takes up pictures of food and LinkedIn takes up details of jobs. A lot of humans do fail to realize that their choices are affected by their interaction, or they are a part of the giant social computer. There are a number of studies that study the direct and indirect impact of social media on humans.

3.2 What is human?

The answer to what humans really are is a never ending discussion across the social sciences and humanities disciplines. In HCI, Ben Shneiderman, a leading researcher, intends to get rid of the term, 'user-friendly', because it presupposes a universal overarching generalized definition of the "user". HCI research, no doubt, has gone deep into resolving this issue by including users of different age groups, cultures, and genders in their research, but without paying adequate attention to the idea that these users are conditioned by the tools that have been using for hundreds of years. The idea is not about inclusivity or diversity, but about being aware of the fact that the users are dynamic evolving entities. The question of what is human is then best left as an open question for the researcher to meticulously probe.

3.3 On the AI critique

In the light of the above insights, let us now look at the critiques that AI has been facing. One of the critiques is that AI would outcompete the humans and slowly make them obsolete. Competition requires a bounded environment and clear metrics to operate. One competes in a 100 m sprint race. Ideas, such as competing for life, simply do not make any sense because there can exist no consensus over the metrics used for such a measurement. In a 100 m race, time and distance are measured and there lies absolutely no disagreement over what each of them means. Competing for life thus is absurd because of the lack of consensus over the metrics. By this logic, the idea that AI is competing with humans (say for jobs) is slightly misplaced. It is necessary to determine what sort of jobs we are really looking at. Is it a simple repeatable low-level job description that can be replaced by writing a few lines of code? Is it a morally complex but low-level job that a machine is not supposed to do? Or is it a risk-laden, loosely defined job that a machine cannot do? Routine, bounded, morally resolvable bureaucratic jobs are ones that can be taken up by machines. These are jobs with succinct job descriptions. Creative, non-bureaucratic and other jobs requiring high agency and a moral calculus are examples of jobs that have loosely or undefined metrics and, by definition, machines cannot participate in.

The other idea of humans becoming obsolete is grounded upon the notion that humans are fundamentally functional in nature. This means to say that the basic human attribute is to be able to function in a given setting (economy, family or others). The word functional means performing a function for fulfilling a certain end. Feeling, experiencing and thinking are not always the means for a specific end, neither are ends in themselves yet fully human activities.

3.4 The emerging HCI–AI

With the new arsenal of insights, where do we stand at the HCI–AI junction? HCI then has not only to deal with a new AI, but also with a new human. As AI evolves, so does the human in a yin-yang recursive manner. One feeds the other. Trying to dissect or disentangle such an arrangement is to undermine the impact of one or the other. A visible implication on the HCI community is that the comforting ideas of human or user-centric need to be reformed, because the idea of human itself evolves with the tools. A physical implication of this is not that a designer must impose tools upon humans, but rather the designer keeps an open and evolving idea of a human or user while designing tools. Humans shape the tools, then tools shape the humans.