Human Agency and AI

[Updated on 4 June '26 with additional paragraphs on mental agency.]

[This article has also been kindly translated into Italian by Francesco Galgani.]

It is one year since my last post, in which I related how I made a comparison of human and machine intelligence on the basis of reflections on stillness. After pulling my paper from consideration for the JSRE special issue, I subsequently had a shorter article on this topic included in Issue 78 of De Numine, the magazine journal of the Alister Hardy Trust.

Since my last post my Academia profile has received dozens of invitations to submit a conference paper, all of them relating to AI. The headline themes include computer science, engineering, signal and image processing, wireless and mobile technologies, database management, big data, Internet of Things, and so on.  Yet, at the same time, I’ve not received any invitations related to my most popular upload by far, which is on Buddhist ethics. This is an indication of the frenzied and scattered attention being given to AI, often at the expense of pretty much everything else.

Mainstream media is reporting increased existential fears, ranging from the immediate (job uncertainty) to long-term implications for humanity.  This is concentrating attention on the distinct qualities of being human, but I find the response from academia tends to take a materialist perspective and gets bogged down in legalities.  Much more on point is Pope Leo XIV’s encyclical Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence.

The Pope's letter dwells especially on the social aspects and the common good, naturally reflecting the concerns of an institution. Here I take a Buddhist perspective, focused more on the individual.  Moving out of stillness, I wish again to highlight six characteristics of human agency:

    • initiation
    • exertion
    • making effort
    • steadfastness
    • persistence
    • endeavouring

This is sourced from the Attakārī Sutta: The Self-Doer, AN 6.38 (an English translation from the Pali on Access to Insight). 

As an illustration, imagine entering the London Marathon. Fast forward to race day and the run itself.

    • initiation: when the starter fires the gun, you respond and decide to start moving
    • exertion: you drive forward and start moving your legs
    • making effort: you accelerate up to your normal running speed and into your standard running posture
    • steadfastness: you maintain a steady gait and pace, start to clock up the miles
    • persistence: you tire and start to ache, but keep going 
    • endeavouring: you ‘hit the wall’, but take on energy fluids, rest a while and then keep going, determined to finish.

You complete the race and glow at the achievement; the sense of personal satisfaction is immense on top of the gladness in raising funds for your charity.

The example primarily concerns physical agency, though in the wider context of preparation and training also involves actions in speech and mind.  These activities are permeated by the pervasive digital environment, involving many online interactions such as registration, making purchases, and seeking inspiration and advice in social media. 

Getting ready for and participating in a marathon is usually highly motivated and focused, strengthening agency and increasing confidence in facing the challenges of life, but is this the case for other aspects of a typically busy lifestyle?  When online or plugged in with digital devices, we need to be especially heedful of the mental aspects and be wary of being swept along.

• Are we making decisions ourselves?
• Are we attending to tasks promptly or putting things off?
• Are we thinking things through?
• Are we evaluating for the long-term as well as short-term?
• Are we prepared to put in the extra hours to solve problems? 
• Are we ready for a power cut?
• Do we have a plan B?
• What have we really learnt?
• Are we able to stop and bring our minds to peace and stillness?
• Are we contented?

Or are we restless, increasingly handing over tasks to some other agent, generative AI, in the hope that it will solve our problems?  With the nature of this technology being all-embracing, there is a potentially huge deleterious impact on cognitive functioning. It could impair human agency to such an extent that we become little more than drones.

Rather than blindly accepting the 'latest and greatest' digital promise, we should always ask: Do we really need it?  Constructively, we can take each of the six characteristics listed in the Attakārī Sutta and simply ask of any proposed technology: Is this likely to enhance or undermine human agency?  To sharpen the focus, ask these questions assuming that the technology is removed after a certain time. What are you left with?

I discuss this further in my short book, Buddhism and Computing: How to Flourish in the Age of Algorithms, and paper, Cultivating sīla Online: the use of Cognitive Interventions in Systems Design.

Two Research Papers on Religious Experience and Artificial Intelligence

I have just drafted two papers: How might Artificial Intelligence be Approached for Research in Religious Experience? (available on Academia or on my own website), and How might Artificial Intelligence Support Research in Religious Experience? (also available on Academia or on my own website).

These papers came about after the Journal for the Study of Religious Experience  put out a call for papers for a special issue on AI. I was initially reluctant to respond as I had (and still have) reservations about the list of topics and areas. However, I was urged to apply by a member of the Alister Hardy Trust, who pointed out that I have studied both computer science and religion. So, I sent off an enquiry, establishing that this was to be the first issue to treat AI.

After a bit of further thought, I submitted the following abstract for consideration:

This paper explores the subject of Artificial Intelligence with a view to enhancing research activities in religious experience.  It starts with a broad examination of AI from a historical perspective as an evolution in notions of computation and in the development of machines, with comparisons made between machine and human intelligence. Further evaluation is carried out on the nature of the systems and theories, using a Buddhist lens, focusing especially on intention, agency and creativity, with references from the Theravada canon.  Observations on stillness draw out profound distinctions, providing indications of what is or is not appropriate in terms of AI expectations and engagement.  From these considerations, a practical application of AI emerges in terms of augmenting human intellect, along the lines envisaged by Douglas Engelbart, a pioneer in the field. Accordingly, initial indications are provided of how the work of RERC may be enhanced by the analysis of its archive of personal testimonies using machine learning to implement techniques in natural language processing.  These may include topic analysis of, for example, subject matter and circumstances; and sentiment analysis of emotional states induced, thereby helping to corroborate and extend existing findings.

The journal’s editorial team accepted the proposal, indicating that they were particularly interested in the possibility of AI assisting with the database of accounts at the Alister Hardy Religious Experience Research Centre (RERC). So I subsequently registered for database access, chose a theme (stillness) as a basis for comparing human and machine intelligence, and proceeded to draft a paper, which I managed to submit by the deadline.

My submission was reviewed and after a few weeks I received some feedback. This made me pause to reconsider and after a few days deliberating I decided to withdraw my submission. I shan’t go much into the reasons, but evidently, I had not conveyed clearly enough the main argument, expressed in the opening pages, plus the various applications of AI drew attention away from this argument to the affordances that AI offers. That is what I wanted to avoid. So, the paper needed to be split into two (the argument followed by the application), whilst keeping the existing order. After the split, I expanded the first paper a little to add a key message: scholars and practitioners of spiritual and religious experience should become more involved in the evaluation of AI especially because they have a fuller understanding of what it means to be human.

It is becoming common for universities to address this question (a Google search over the Oxford University domain), quite often in interdisciplinary panels, but I find they do little to affect the general direction of travel, which continues with rapid extensions and expansion of development, albeit with some regulatory hurdles? One of the underlying problems is the repeated pairing of human intelligence and machine intelligence, as though they are comparable side by side. Other terms have built up and established anthropomorphic language (machine learning, deep learning, and so on), which compounds the issue and leads to tacit acceptance.  

There are, in my view, more suitable alternatives. Whilst working on the KEPLAIR project, I was made more aware of the contrast between neural networks as a ‘black box’ technique and transparent techniques based on logical reasoning. In this connection, I was introduced to Douglas Engelbart’s work on Augmenting Human Intellect. Rather than frame questions around AI, we frame them in terms of enhancing humanity.  Accordingly, this framework underpins the second paper, which is more conventional in that it makes suggestions of machine learning applications to support research into religious experience, particularly the RERC accounts. This is largely in the area of hermeneutics (for which Voyant Tools is well-suited), though the requirements on data protection encourage more unusual possibilities.

The papers are currently unpublished drafts. I’d welcome suggestions for open access publication.