Part 3 – Speaking of futures: Presuppositions

Part 1 – Speaking of futures: Story-capsules  | Part 2 – Speaking of futures: Que será, será

Any given subject can be framed in many ways – as the arts make manifest – yet all too often, we resort to the reductive framing of binary oppositions: good vs bad, happy vs sad, promise vs threat. Even UN Secretary-General António Guterres offers us the choice of ‘reform or rupture’ for the UN. One of the most entertaining instances of this binary framing (cited in my previous blog post ‘Part 2 – Speaking of Futures: Que será, será‘) is Noah Sweat’s ‘whisky speech’ in which the Mississippi judge depicts whisky first as a scourge and then as a blessing and is equally convincing about both. I suggested there that our current attitudes towards AI indicate a similar ambivalence: is it good or bad, friend or foe, saviour or destroyer of humanity?

This week, I consider how this ambivalence is brought about by a very particular framing device – presuppositions. I illustrate my argument primarily with magazine covers such as TIME and The Economist.

Presuppositions are one of the most covert forms of framing. The statement ‘Play it again, Sam’ presupposes that Sam has played it before. Significantly, whether I say ‘Play it again, Sam’ or ‘Don’t play it again’, the presupposition that Sam has already played the tune still holds. Thus, presuppositions have the unique characteristic of holding under negation: they are like a virus that infiltrates a proposition and cannot be dislodged, no matter whether one affirms or negates the proposition. For example, ‘Did you or did you not stop smoking?’ presupposes that you have been smoking. ‘Do you know what your problem is?’ and ‘Don’t you know what your problem is?’ both presuppose that you have a problem.

Why focus on presuppositions? Precisely because they are covert, omnipresent, and hard to negate. Indeed, more often than not, the best way to challenge a presupposition is not to negate it but to reframe the narrative. And that is exactly what we see in the TIME covers above. ‘The End of Humanity’ presupposes that the end of humanity exists (with the implied promise that the subject will be covered by the relevant article in the magazine). The negation test yields propositions such as, ‘The end of humanity will not be AI-generated, but man-made’, in which both options still assert that ‘the end of humanity’ is a given. ‘Advancing Humanity’ simply reverses the message by substituting the pessimistic ‘end’ with an expression of ongoing progress.

Similarly, in the iconic 1989 cover ‘Endangered Planet’, the presupposition is that the planet is endangered (a statement that we can then elaborate by means of ‘not because of X but because of Y’). The visuals furthermore suggest that we are barely holding it together with strings in a post-apocalyptic climate. In the AI-generated variant, the strings have been replaced by a web of connectivity, the image sports blue-planet colours, humans are shown to be in control and the text–image pairing implies that ‘bringing AI everywhere’ is a good thing.

This most simple form of reframing, the reversals of the value ‘bad for us’ into ‘good for us’ is, I believe, at the heart of our ambivalence towards AI. We keep being fed frames both for and against, whether these are in the form of images, captions, or arguments, and unless we have a critical-thinking mindset, we are likely to jump straight onto the bandwagon that resonates with our gut response: attraction and happy anticipation, or aversion and fearmongering. When we run with our biases, we rarely stop to question the perspective and moral judgment that any given frame promotes. Magazine covers excel at grabbing our attention and fuelling our biases by means of presuppositions.

How then can we recognise presuppositions in order to guard against their subliminal influence? In what follows, I identify some common presupposition triggers to show how they help to frame an issue. The most prevalent triggers are definite and possessive markers. ‘AI’s new frontier’, ‘Big Tech’s supersized ambitions’, ‘the risks of generative AI’, and ‘the end of the social network’ all presuppose that what is being referred to exists. As does ‘the king of France’s beard’.

Questioning such presuppositions would involve:

1. Fact-checking: Is the conflict in Ukraine really the first AI war? What percentage of AI-driven weapons, information gathering, and tactical planning qualifies a conflict as an AI war?
2. Value-checking: What is the emotional load conveyed by connotations and other story-capsules? Are Big Tech’s ambitions really unrealistically supersized (a term which carries negative connotations coupled with images that caricature superheroes) and who judges them to be so? Is the social network really going to end (a sad demise, to judge by the emoji) or will it be replaced by a better version (as Meta et al. assure us)? The image–caption coupling of magazine covers often reinforces its emotional appeal.
3. Narrator bias: Who is framing the issue in this way and why? Is the commercial imperative to stop page traffic and increase sales encouraging the design of more ‘arresting’ frames than the arguments promoted within the magazine? Is there a knee-jerk reframing reflex that favours binary thinking?

Temporal clauses, such as ‘after the launch of large language models (LLMs)’ and ‘before the advent of artificial general intelligence (AGI)’, presuppose that LLMs and AGI exist and were either already launched or are about to land. ‘Once general AI takes over’ presupposes it will, as does ‘when it exterminates us.’

Change-of-state expressions presuppose that the previous state existed: ‘is born’ presupposes ‘was not’ before, which leaves space for ‘gestating’ or just ‘non-existent’. ‘Wakes up’ presupposes ‘was asleep’, ‘new’ presupposes ‘old’, ‘breaking through’ presupposes that progress was ‘blocked’ before, and the term ‘revolution’ itself denotes a change of state.

Comparatives and superlatives such as ‘bigger’, ‘better’, ‘more intelligent’, ‘the first’, and ‘the most recent’ presuppose that the other/prior entity or state (‘smaller’, ‘less good’, ‘less intelligent’, and ‘not so recent’) also exist(ed). Comparatives often overlap with change of state expressions and with temporal clauses, as in the phrase: ‘in a time of increasing uncertainty, complexity, and accelerating pace’. These category distinctions are not important, what matters is that we recognise when presuppositions and assumptions are inserted into the discourse by means of everyday expressions.

Implicative verbs are another presupposition trigger and include expressions such as ‘to manage to’ (presupposes that you were trying and implies that you succeeded), ‘it happens that’ (presupposes that chance is involved), ‘to fail at’ (presupposes that you tried), ‘to plan to’ (presupposes premeditation), ‘to fight for’ (presupposes competition for victory), ‘to scramble for’ (presupposes competition over a scarce resource), ‘to take control of’ (presupposes that one does not currently have control but should). ‘To take back control’ adds an iterative meaning to the expression.

Iteratives similarly act as triggers, as we saw in ‘Play it again, Sam’ and as is evident in ‘Make America great again’ and the UN 78th General Assembly’s focus on ‘rebuilding trust and reigniting global solidarity’. Finally, the question ‘how’ presupposes that the action under consideration can be undertaken and directs our focus onto the method: ‘How should we teach AI right from wrong?’, ‘How do we measure the risks?’, and ‘How will we combine the real and the digital worlds?’

My personal favourites are factive expressions. These presuppose that the proposition which follows is a fact, even though it may be a fabrication. ‘I realise you are worried’ presupposes that you are indeed worried. So do: ‘I’m sorry that you are worried’, ‘The disadvantage of your being worried is…’, ‘It’s a pity that you’re so worried’, and ‘It’s strange that you’re so worried.’

Factive expressions are such a familiar feature of everyday language that we rarely recognise how they spin what is presented into a given. It is fascinating to note that factives far surpass non-factive expressions (those which do not presuppose the truth of what follows). In the case of non-factive expressions such as ‘assume’, ‘believe’, ‘suppose’, ‘claim’, ‘conclude’, and even in certain uses of ‘know’, the speaker is expressing their belief rather than any certainty. As a result, the proposition that follows may either be true or false. In the following paired sentences, (a) is a factive and (b) is a non-factive verb:

1. (a) I am glad that you are reading this.

    (b) I assume that you are reading this.

2. (a) Are you aware that the future is uncertain?

    (b) Do you believe that the future is uncertain?

3. (a) It is sad that he has left.

    (b) It is said that he has left.

The following are only a few examples of all the existing factive expressions. This is an open class which can readily be added to:

• Realise, remember, recognise, regret, discover, learn, understand, accept that, be aware that (known as ‘cognitive factive verbs’)
• Be sorry that, be proud that, be indifferent that, be glad that, be sad that (known as ‘emotive factive verbs’)
• Be a fact that, be worth it, be strange that, be odd that, be regrettable that, be lucky that, etc.

Why do I love factives so much? Because they pop up absolutely everywhere in our language (no matter what language we’re speaking) and they are essential in creating the narratives that we spin. The most in-your-face factive expressions are: ‘it is a fact that’, ‘we must remember that’, ‘we must never forget that’ – all expressions that pepper the speeches of politicians. We often find factives in the opening lines of novels too.

‘It is a truth universally acknowledged, that a single man in possession of a good fortune, must be in want of a wife.’

– Jane Austen, Pride and Prejudice

It is a truth according to whom? Certainly not according to any young millionaire of my acquaintance, but according to Mrs Bennet, the mother of five girls, whom it is her life’s mission to marry off. Whether we apply the ‘Mrs Bennet test’ to matrimony or artificial intelligence, it is always worth asking of any factive expression, or indeed of any framing device: Who says so and why? To what extent is it subjective? Can the proposition be fact-checked? Is it emotionally loaded? In what ways is it disputable? And what other pertinent truths or facts does this particular proposition obscure?

To conclude, what do we learn from applying the Mrs Bennett test to the subject of futures literacy?

1. Who is inviting us to envisage possible futures and for whom do they speak, with what bias? Is it a rich-country perspective? Is it an international organisation agenda, and as such, is it the beginning of a new era of reform or just a passing fad over the latest theory on the block? How might the Global South perspective differ, or the perspective of religious institutions, many of which have their own views of the forces that shape the future?
2. Fact-checking would require us to ask whether we are indeed at a turning point, as all the rhetoric proclaims, or whether we are simply in a period of accelerated change brought about by a technological paradigm shift, as we have experienced previously. We would also want to know what is the remit of the complexity model with its emphasis on emergence? Does it apply across the board or just to some aspects of human behaviour and to some aspects of the human–AI interface. And more fundamentally still, is AI really intelligent, or sentient, or (un)ethical, or otherwise capable of the human attributes that our tendency to anthropomorphise leads us to believe?
3. Understanding the emotional load of any futures frame involves, at one level, attending to the connotations, metaphors, logical fallacies, presuppositions, and other story-capsules associated with any given frame, as outlined in this trio of blogs. But at another level, the now-or-never urgency associated with the project of futures literacy creates a sense of anxiety or FOMO (Stefano Baldi) that also colours our view. Whether ‘we are driving change or change is driving us’ (Paula Middleton) also needs to be addressed.
4. Finally, we might ask: ‘What else could be said about the future, what has been omitted?’ There is no discussion of religion, for instance, at the risk of overlooking the billions of people for whom God is a more significant agent than man or AI. Similarly, in the terminological conflation surrounding AI, we may need to make clearer distinctions between different types of intelligence, different ways of thinking, of learning, and of communicating. These distinctions exist between humans, and are all too readily ignored. How much more is this the case between humans and AI?

Even our current view on complex emergent systems has precedents:

‘We insist that our concept of perspective is the total antithesis of all static perspective. It is dynamic and chaotic in application, producing in the mind of the observer a veritable mass of plastic emotions’

– Carlo Carrá, futurist artist

My aim at the start of this trio of blogs was to explore how our use of language influences the way we visualise the future, or indeed the present. In particular, I sought to expose how linguistic resources that are available in all languages may influence the way we construct narratives and respond to them. My claim has been that at the word and morpheme level, the story-capsules we find in connotations and metaphors very evidently colour our perception. At the level of larger-than-word narratives, whether these be logical fallacies, analogies, cautionary tales, or science fiction, we are similarly influenced by the emotional and moral content of frames. We seem to be particularly primed to respond to ‘friends vs foes’ and ‘safety vs danger’. Whether we err towards fatalism or strive towards agency, these inclinations are believed to be emotionally driven and can readily be activated with narrative frames which amplify our own biases.

Finally, in this blog on presuppositions, I have looked at the binary nature of much of the discourse on AI and possible futures, and have suggested that this may in part be due to our use of presuppositions, as illustrated by magazine covers. Because presuppositions are endemic in everyday language and are hard to dislodge, they are often dealt with through binary reframing. In an age which promotes the communication of immediately accessible image–caption coupling over word-based argument, that binary approach helps create pro and contra camps, echo chambers, and further bias amplification.

My hope throughout these blog posts has been to heighten awareness. We are all native language speakers, yet we are not always cognisant of how language works and how it works on us. If at any point during your reading you sat back and thought, ‘Oh wow, I didn’t realise that!’ and followed that thought up with, ‘Ah yes, I recognise a factive when I see one!’, then I am happy.