The Questions We Never Ask: A Christian Relational Approach To AI
“The greatest trick the devil ever pulled was convincing the world he didn’t exist.”
Charles Baudelaire’s words were popularised in the 1995 film The Usual Suspects, in which they are quoted by a low-level conman who turns out to be a criminal mastermind.
The point is that if we do not believe in an evil influence, whatever form that takes, it can operate unchecked.
AI raises the possibility of an even more insidious and dangerous trick being pulled. Not that we don’t believe a problem exists – that we consider it and dismiss it – but that we never think to ask the question.
New technologies typically allow us to do things more easily or effectively than before, or occasionally to do things that were not previously possible at all. Due to the nature of AI, some of its most serious risks are harder to spot than for other technologies.
AI enables us to circumvent activities that have, up to this point, only been achievable through human thought, feeling, and judgment. Many of the activities we use it for are fundamentally emotional and relational. This is different to previous technologies, like calculators and computers, which automated tasks. Instead, AI automates selfhood – or produces a convincing facsimile of it.
Selfhood is, of course, critical to Christianity. We choose our faith, and we make daily decisions that reflect and contribute to that faith, forming habits and character.
In Galatians 5, Paul writes: “But the fruit of the Spirit is love, joy, peace, patience, kindness, goodness, faithfulness, gentleness, and self-control. Against such things there is no law.”
All of these fruits are relational qualities, developed over time and through conscious effort. Uncritical use of AI risks short-circuiting that development. The fruits of the Spirit aren’t performed or executed: they are grown.
From Narrow To Generative AI
Artificial Intelligence had always been the stuff of science fiction: sentient computers like HAL in Arthur C. Clarke’s 1968 novel, 2001: A Space Odyssey. For decades, imminent breakthroughs were forecast but never materialised. When the breakthroughs finally did take place, they happened almost overnight.
Forms of AI have been used for many years. So-called “narrow” or “weak” AI – algorithms specialised for a single task – powers many online services, including the recommendation engines that curate our social media feeds. These have already been recognised as driving addictiveness and undermining our ability to concentrate by rewiring the brain to seek short interactions and sensitising us to distraction. How can we maintain relationships, with God or anyone else, when we cannot listen or pay attention for more than a few minutes? This and other problems have been covered extensively elsewhere. [1]
The more recent “generative” AI is different, and far more powerful. As the name suggests, generative AI creates content: text, audio, images, video, and more. This is a distinct technology to narrow AI and comes with a separate set of risks and benefits.
A significant milestone for generative AI was ChatGPT’s public release in November 2022. This was the point at which anyone could query OpenAI’s sophisticated AI chatbot, gaining human-like answers on almost any topic they wanted. Within two months, ChatGPT had gained over 100 million users, making it the fastest-growing app in history at the time.
AI has continued to develop quickly ever since, with the release of new and more reliable models. The answers AI gives are not always perfect, or even accurate but AI tools are now embedded throughout online services and the way we work.
Some of the problems of this form of AI are already well-recognised. AI-generated images and videos blur the line between the real and the imaginary. “Fake news” becomes far more convincing in the era of generative AI. And, of course, AI models each have biases in terms of their ethical frameworks and perspectives on controversial political and social topics.
With the existing recommendation algorithms, this gives rise to an even wider and more carefully-tailored set of curated realities: a combination of real and generated content, served via algorithm – determining what we see, and therefore what becomes thinkable. This problem, though, largely predates generative AI. Additional concerns typically focus on AI’s energy use, or security threats.
For Christians, generative AI introduces risks that are harder to name than these, and harder to spot precisely because nothing appears to go wrong.
In Frank Herbert’s Dune series, “machines in the likeness of a human mind” were banned. The reason was not that they were malevolent or posed a direct threat. Instead, “Humans had set those machines to usurp our sense of beauty, our necessary selfdom out of which we make living judgments.” [2]
This is the immediate and often overlooked risk of generative AI: that we hand over work to it that is a core part of our character formation and spiritual development.
The Relational Risk
Christianity is a relational religion. [3] When Jesus summarised scripture, he did so in terms of love for God and love for neighbour (Matthew 22:34-40). Every law and command in the Old Testament governs one or more aspects of a relationship. The Jubilee Centre has always articulated the meaning of life in terms of relationships.
This is therefore also the Jubilee Centre’s model for what it means to be human. We are made in the image of God (Genesis 1:26-27). As a trinitarian being, God himself is inherently relational, and humans are relational to their core.
Life online, and particular use cases for generative AI, raise the idea that we might live more and more of our lives “relating” to something that is human-like in certain aspects, but that is not human. For tech companies like Meta, this is a stated ideal: that AI chatbots will become an extension of our friend networks.
This is dehumanising, because we are relating to something non-relational. At the same time, many of the dimensions of relational proximity (Directness, Continuity, Multiplexity, Parity, Commonality) [4] are mimicked, making the AI-human relationship a high-quality forgery of the real thing. While this is not idolatry in the strict sense, in that we are not worshipping a false god, in the worst cases it could be considered “idolatry-adjacent”, since we are exchanging something made in the image of God (another person) for a relationship with something that is not.
It’s easiest to understand this via an example. Over one in three people in the UK have used an AI chatbot for mental health support. [5] This includes both general-purpose platforms like ChatGPT and Claude, as well as purpose-built mental health services, which offer support in the form of techniques derived from mindfulness and CBT. Effectively, people are using AI as a therapist.
There are good reasons for this trend, including difficulty in accessing mental health services, reluctance to talk to friends/family, and ease of access – it’s fast and convenient to use an app, while an appointment on the NHS might take many weeks to come through, if at all. At the same time, many users (rightly) feel uneasy about using a “therapist” who does not, in a relational sense, exist. There are also concerns around privacy, inaccurate or harmful advice, and lack of regulation and accountability.
Still, there are a number of popular and growing uses for AI that follow a similar pattern:
Companion apps, friends and even romantic partners
Tutors
Personal assistants
Parenting support
Prayer, confession and spiritual guidance
In each case, AI replaces a human relationship. The other party feels like they are known and seen but are not.
There is a legitimate question about how valuable these services are, even with this serious relational caveat. An AI app may still facilitate effective therapeutic work, especially when there are no alternatives. When someone chooses to use an AI therapist, they recognise their need to change and grow, and are actively engaging in that process. People may learn to relate better to one another by using AI to acquire new skills and insights – just as they might through reading a self-help book.
An arguably greater spiritual and relational risk occurs because it is far too easy to use AI to avoid engaging emotionally and relationally – and this can happen seamlessly, quietly, and out of sight.
The Spiritual Risk
In the examples above, someone ends up interacting with a party that doesn’t exist, at least, in conventional human relational terms. They hand over the role of therapist, companion, or tutor to AI, allowing a program to replace the emotional processing of another person.
A more insidious problem occurs when we hand over our own emotional processes to AI. This is surprisingly easy to do.
AI answers our questions immediately, fluently, and very often confidently. Sometimes it gives bad or incomplete answers, though the models are rapidly increasing in sophistication. Its answers are generally well-reasoned and make perfect sense, mixed with unending patience and encouragement. It is this apparent competence that poses the greatest risk.
Relationships and spiritual growth are not simply a matter of getting the right outputs – of acting a particular way of being. And yet AI can be used to carry out tasks that typically require patience, uncertainty, and emotional reasoning.
Let’s say you need to write a sermon but aren’t sure where to start, or perhaps you’re just busy. Generative AI could be helpful in getting the job done. It might even come up with a better sermon than you could, and certainly in a far shorter time.
You then give this sermon in church. Or to put it a different way, you perform it, like an actor delivering a script. You may get some benefit from the sermon, perhaps the same benefit the congregation gets. What you don’t get is the benefit of wrestling with the questions.
Typically, writing a sermon would involve reading the relevant biblical texts and commentaries, prayer, discussion, and careful thought – all over a period of time. Any personal spiritual and character development that might have come from preparing the sermon is lost when you hand that task over to AI, even if the end result looks very similar. You might not even notice what you’re missing, because what you lose out on is the opportunity to develop.
Of course, the reality would probably look slightly different. You probably wouldn’t use AI to write the sermon from start to finish. But you might use it to generate ideas around the topic and perhaps suggest points of application.
You might be pleased with the result. The AI-assisted sermon might even be more beneficial to your congregation than one you wrote entirely yourself. Over time, though, using this process unthinkingly could hollow out your own spiritual development and capacity to interpret scripture.
The same is true of other activities where difficulty is integral to the process. That difficulty is not an optional extra or an inconvenience, but part of the mechanism of decision-making. As the famous quote goes, “It’s about the journey, not the destination.” [6]
Take another example. Let’s say you’ve had an argument or disagreement with a friend. You’re tired, upset, and not sure what to do about it. You don’t want to discuss it with other mutual friends, since that would put them in an awkward position and they’re unlikely to be impartial.
So you turn to an AI app. It’s always available, is never judgmental, never gets impatient or annoyed with you, and no question is too trivial or embarrassing for it to answer.
The app gives you advice – maybe good advice – about how to handle the argument. Instructions about taking time to cool down, owning your part in what went wrong, reaching out (including suggested wording), listening to your friend, and figuring out what fixing the problem looks like in practice. [7]
You follow the instructions, it works, and the friendship is saved. It was actually fairly easy and painless. You avoided many of the mistakes you might have made without AI assistance.
At some point, it might cross your mind that it felt just a little too easy. Probably, though, you’re just grateful the situation has been handled with a minimum of fuss and you can get back to how things were. Following a set of well-crafted steps is simpler than engaging with messy reality and uncertainty.
One of the things that is missing, or much reduced, is discomfort. You never really needed to sit with the emotions for long, to take the time to understand and process them. Like the sermon example, AI is a way to avoid doing not just intellectual work, but character work.
Another thing that is missing is authenticity, as well as agency. You did the “right” thing, in the sense that it addressed the immediate problem effectively, but was it really you? Or have you essentially executed a script that leads to the desired outcome?
Consider how many passages in the New Testament emphasise the importance of Perseverance, and why. Just a few include:
“We also glory in our sufferings, because we know that suffering produces perseverance; perseverance, character; and character, hope.” (Romans 5:3-4)
“Let us not become weary in doing good, for at the proper time we will reap a harvest if we do not give up.” (Galatians 6:9)
“Consider it pure joy, my brothers and sisters, whenever you face trials of many kinds, because you know that the testing of your faith produces perseverance. Let perseverance finish its work so that you may be mature and complete, not lacking anything… Blessed is the one who perseveres under trial because, having stood the test, that person will receive the crown of life that the Lord has promised to those who love him.” (James 1:2-4, 12)
This is not talking about needless perseverance, or self-imposed pain. It doesn’t mean we should deliberately seek out hard situations. But when difficult situations come along, as they inevitably will, the challenge is to engage with them at a deep level, not to avoid the emotional and relational work that is a part of character formation.
The example of the argument with a friend also shows how easy it is to use AI as a tool to manipulate people – to find the right form of words and actions to bring about a desired outcome. That also applies to workplace relationships and other situations where you want to “get the best” out of a given encounter.
Jonathan Burnside identifies this kind of manipulation as a modern-day Western parallel to witchcraft. [8] The essence of witchcraft is a relational void: “It expresses our failure to enter into a relationship with God, to make discoveries about the spiritual and the material worlds in partnership with him, and to fully trust him with our lives and futures… we are each guilty, at some level, of trying to manipulate people and events to bring about what we want, without trusting fully in God. We can all live in a way that gives us a sense of power and control, whilst denying God’s control and power over the minutiae of everyday life.”
Using AI to shape our relational interactions treats relationship itself as an optimisation problem. In doing so, it reduces both parties: ourselves to technicians, the other person to an outcome.
It’s worth repeating that the broader argument is not about deskilling – at least, not in terms of “hard” skills. That has always been a common criticism when new technologies arise. Calculators mean we lose the skill of using a slide rule for trigonometry; SatNavs mean we no longer learn to read maps. The outcome (calculating an angle, finding our way from A to B) is still possible, but the way we achieve that task changes. Depending on the nature of the task, the loss of a skill – or more seriously, the cognitive capacities that lie behind that skill – can be a problem. While the world has moved past the need for slide rules, when the SatNav fails many of us no longer have the backup map-reading skills (or mapbooks) to figure out the journey.
The problem with AI is not simply that we lose the ability to do something. It’s that we lose what we have the ability to become.
The Habitual Avoidance Of Difficulty
The human brain is hard-wired to seek economy, reducing unnecessary effort to a minimum. When we know we can find information somewhere external (like the internet) in the future, we don’t make the same effort to retain it ourselves. “The Internet has become a primary form of external or transactive memory, where information is stored collectively outside ourselves.” [9] We understand this anecdotally: we’ve all reached for a calculator to do a sum we could do in our heads or on paper with a little extra time and effort.
This mechanism appears to be self-reinforcing: if we use the web to answer one question, we’re more likely to use it to answer subsequent questions, even if they are easier than the first. [10]
While AI is still at too early a stage for studies to have been undertaken, it’s a reasonable assumption that the same dynamic will hold. The existence of sophisticated AI services shapes our decision space – the sum total of choices we have at a given time – and we are likely to take the path of least resistance. We will increasingly offload cognitive and emotional processing if the opportunity exists.
Take again the example of the friend with whom you have a disagreement. AI gives you an effective solution, but one that optimises for the superficial state of the friendship rather than long-term personal and relational development.
If you press the AI about these dangers, it will agree with you and provide further insights that could be genuinely beneficial – but that is only something that will happen when you already know it is important. In the general scheme of things, you will come to the AI app with a problem, and it will solve it for you. It may not even occur to you that you’re missing something important, precisely because the outcome is so slick and effective.
This is the mechanism alluded to in the Baudelaire paraphrase at the beginning of this article. Not that we dismiss the threat that AI poses, but that the nature of AI means we never even see it coming. We quietly give up opportunity after opportunity to develop spiritually, emotionally, and relationally, while at the same time securing better short-term outcomes than we ever would have done ourselves. The answer might be “right”, but the process that leads us to it – and builds character – is entirely missing.
Over time, we become less capable of the kind of thoughtful consideration and applied spirituality that is fundamental to everyday life as a Christian.
Application
Where does all this leave us as Christians living in the era of AI?
Human history has been a story of constant technological development, and there’s nothing new about new technologies impacting our society and behaviour. However, there is something unique about generative AI, because it is capable of automating selfhood: of undertaking emotional and relational as well as cognitive processes that are fundamental to what it means to be human. This raises the prospect of a growing percentage of the population abdicating emotional processing and decision-making to an external entity.
History of Technology professor Melvin Kranzberg famously coined Kranzberg’s Six Laws Of Technology. [11] The first law is:
“Technology is neither good nor bad; nor is it neutral.”
Kranzberg’s point was that any technology is expressed within a specific context and set of circumstances, by humans with stated or unstated values and agendas. The same piece of technology deployed in two different ways could have two very different sets of impacts.
The best solution to a technology that poses risks is generally not to avoid all use of any application of that technology. It’s to go into it with our eyes open and make conscious decisions about how we engage with it.
Given the benefits and risks of AI, there are at least three ways in which we can use it positively.
1. Create friction when necessary
The danger is that we outsource emotionally and relationally difficult processes to AI, thereby foregoing the opportunity to develop ourselves – and ultimately, reducing our capacity to engage with hard situations.
One answer to this is deliberately to use AI as a sparring partner. Get it to ask us hard questions, and to interrogate us to make sure we are not making life too easy for ourselves.
This may involve asking for critiques on certain matters, without (initially) asking for solutions. For example, you could use AI to “red team” a proposed response to your friend you wrote yourself: arguing against your position, finding potential holes in your thinking, and so on. The aim is specifically using AI to grow, even if that is painful for you, rather than to avoid that process.
There may equally be times when the right, and hard, thing to do is avoid using AI altogether.
2. Use AI as a learning tool
The risk of “transactive memory” is that when we know we can find information externally, we are less likely to remember it. The extension of this in the context of AI could be called “transactiveprocessing”: loss of the ability or inclination to reason emotionally and cognitively if we can secure that service elsewhere.
However, AI can be an excellent learning tool, if we use it correctly. Again, this involves consciously working a little harder in the short term, for greater long-term development.
In the example of the sermon, you might use AI to assist your work, but take the time to understand and internalise the principles involved, not just use it as a replacement for your own efforts. Instead of using it to create ideas without you, use it collaboratively to generate and test them and acknowledge to the congregation how AI has been used.
Now, you’re using AI to sharpen something you’ve already done: as a whetstone, rather than as the blade itself. This only happens after you’ve carried out the hard work, wrestling with the text and the empty page, not instead of it. At no point is your own judgment bypassed.
This process also involves pushing back against the AI’s input, actively engaging with it and seeking to challenge it, as well as yourself.
3. Protect, don’t replace, relationships
Lastly, we are relational creatures, made in the image of the relational God. Character development occurs within relationships. Protect those relationships.
Every use of AI should come with the question: Does this further a relationship, or detract from one?
Sometimes the relational deficit may be a necessary cost, as with AI-based therapeutic tools where in-person alternatives are inaccessible. Nonetheless, in order to make that decision in an informed way, we have to identify the relational deficits that may be involved in turning to AI in any given situation.
Short-term convenience in what we do means we forego part of the process and conditions by which we become. The fruit of the Spirit, the way its aspects are formed, and the New Testament’s emphasis on perseverance all point to the same truth: that messy, difficult reality is what forges us as Christians. By avoiding that, we become less human, and a little less of the image of God.
Conclusion
Many technologies have changed the way we live and relate. AI is something different: technology that automates not just a task, but the self. That comes with enormous risks, spiritually and relationally.
The ability to mimic selfhood raises the possibility of “relationship” with something that is fundamentally non-relational. Since humans are made in the image of God, and relating to other people is part of that innate nature, substituting a relationship with a person for a “relationship” with AI is dehumanising.
However, the worst risk is that we outsource emotionally difficult tasks to AI, allowing our own capacity for relationship to atrophy, and short-circuiting our spiritual development.
While AI does hold the ability to further our character and relationships if used carefully and deliberately, a high degree of judgment and caution is necessary. To return to the Baudelaire paraphrase at the beginning of this article, the greatest trick the devil ever played is not convincing the world he never existed. It’s ensuring the question is never asked.
The ability to sit with emotional and relational difficulty is what forms us as Christians. We cannot allow any technology to make us less Christian, less human, and less fully the image of God.
Guy Brandon, MPhil and PhD in Old Testament theology
The views and opinions expressed below are those of the author alone and do not necessarily reflect those of the Jubilee Centre or its trustees.
[1] Including Guy Brandon, Digitally Remastered: A Biblical Guide To Reclaiming Your Virtual Self (Muddy Pearl, 2016).
[2] Frank Herbert, God Emperor of Dune (1981)
[3] Michael Schluter and John Ashcroft, Jubilee Manifesto: A Framework, Agenda and Strategy for Christian Social Reform (IVP, 2005), chapter 2.
[4] See John Ashcroft, Roy Childs, Alison Myers , and Michael Schluter, The Relational Lens: Understanding, Managing And Measuring Stakeholder Relationships (Cambridge University Press, 2016).
[6] Widely attributed to American philosopher Ralph Waldo Emerson.
[7] This is a summary of instructions provided by AI platform Grok, in answer to this question.
[8] Jonathan Burnside, Covert Power: Unmasking the world of witchcraft (Cambridge Papers, 2010). https://www.cambridgepapers.org/covert-power-unmasking-the-world-of-witchcraft/
[9] Sparrow, Liu, and Wegner, Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips (Science, 2011). https://www.science.org/doi/10.1126/science.1207745.
[10] Storm, B. C., Stone, S. M., & Benjamin, A. S. (2017). Using the Internet to access information inflates future use of the Internet to access other information. Memory, 25(6), 717–723. https://doi.org/10.1080/09658211.2016.1210171
[11] See https://www.cs.ucdavis.edu/~koehl/Teaching/ECS188/PDF_files/Kranzberg.pdf
