I’ve been following the debates around AAC and artificial intelligence (AI) for several years. During this time I have acknowledged that authenticity of authorship around the words we speak is an important topic for any AAC user.

At the risk of opening up a can of worms, some of the things I have tussled with are:

Why question authenticity of authorship

If an AAC user suddenly starts to say more, how do we know what they are saying is really ‘them’, reflects their thoughts and true wishes?  The concerns being floated are that if a user begins to respond more quickly, provides fuller responses, uses new or novel words, then we should question authenticity of authorship.

When we don’t question authorship

Yet if this were a typically developing child and they started stringing together two or more words, we would be delighted with their progress. We would support them by putting their short utterances into longer sentences; e.g. ‘bye-bye Nanny’ might be recast by a parent saying ‘Yes, Nanny has to go home now, let’s say bye-bye Nanny’. By the time a child is constructing 4 and 5-word sentences, we might be modelling, expanding and developing what they are saying. We do this by either repeating or rephrasing their utterances in a different way, e.g. ‘the cat ran over there’ might become ‘oh yes, the black cat has run over the road, I wonder where he is going?’.  Parents, older children, teachers and others play a vital role in supporting young children in developing their language and communication skills.

What is it that makes us suspicious?

I’m battling thoughts around if a speaking child said the same words an AAC user outputs, would we accept it as typical development and assume their competence? Or because someone is using AAC with an AI interface, are we holding them to a new standard, different to their speaking peers? Are we forgetting to presume competence and assuming it is the AI speaking, and clearly cannot be the person?  Suddenly, it seems people are suspicious of the AAC user’s ability, asking them to prove their thoughts and justify their vocabulary.  So is this a problem of authenticity, or is it an issue linked to ableism?

AI scaffolds everyday interactions

It seems we are moving towards accepting that AI can provide a scaffold to speaking people and assist in writing an email, preparing a presentation, creating a speech, and so much more.   A search of the net shows AI is being used to teach children online, and teachers can use it for lesson plans and marking work.  Without doubt, a learner can access information on any topic, but the concern is still that learning needs to be interactive and not spoon-fed. AI has not removed agency in these examples; it has supported and often, invisibly, scaffolded the process. Providing the ‘author’ checks their work and makes sure it is personalised, we applaud their resourcefulness for saving time, stepping outside their comfort zone or learning a new skill. What we don’t do is stop them, decry their work or assume it isn’t of their own creation.

Learning to use AAC

My question now is, how does a child using AAC interact?  My thoughts are probably in a more structured way than a typically developing child.  Using AAC means having the resources available and having a communication partner to model and interact with.  We currently allow AAC users access to word prediction, possibly with a restricted vocabulary that is age and stage-appropriate.  Best practice says we model and interact during each conversation to support learning; communication is a 2 way process.

Should we be measuring authorship or agency?

Rather than using authenticity of authorship, questioning the message content as the measure of success should we be looking at the factors that demonstrate authenticity of agency?  By this I’m suggesting recognising the AAC user is intentional in both their communication and the interaction, consistent over time, can challenge their communication partner by stopping a discussion; correcting the words selected or changing topic as needed. If an AAC user can demonstrate agency then however they conveyed their message we need to accept it as authentic.

Developing authenticity of agency

Agency for most children is demonstrating the ability to make choices and take action. This is not always an independent action but is frequently scaffolded and enabled by opportunities afforded in the immediate environment and by wider cultural norms. These are nearly always underpinned by interpersonal relationships in formal and informal learning.

Invisible versus invisible scaffolding

Currently, we seem to be suggesting that every time we use AAC with some form of prediction then AI will be putting words into the AAC user’s ‘mouth’.  What we are observing is visible scaffolding, this is laying the various processes of communicating open for any communication partner to see.  Is this wrong or just a different way of learning? With a speaking child we don’t question these environmental and cultural prompts or supporting frameworks.  A good communication partner doesn’t just accept what they see or hear, they clarify and probe for more information to create a shared understanding.  Is the issue then one of most communication partners not being able to evaluate and support agency and intent rather than one of judging authorship?

Does AI predication work?

What happened to the large language models that some AAC resources previously used to support natural selection and recency of use?  The resources that only predicted what the user had previously said?  When the word banks are built on earlier outputs, vocabulary grows with the user. Clearly, this is easier for a text user who can spell any word they want, and this is exactly why teaching literacy to an AAC user empowers them.  But even with symbol systems, when the vocabulary is extended age appropriately and by school topic, then a child needs to know the word or symbol to select it.

Should we apply the same standards for all learners?

Learning isn’t demonstrated by just selecting a symbol or spelling a word. All developing communicators will try to use words they have heard in their own conversations, not always correctly in context or pronounced as we would expect. Yet, we don’t stop them, instead we assume they are trying, making progress and developing ‘typically’.  We don’t become suspicious and conclude they are making a mistake or don’t understand their own mind. So how do we apply this same knowledge and expectation to an AAC user as they develop?

How do we train communication partners now?

At this juncture, should we be questioning the way we train communication partners for children?  Rather than accepting an AAC user’s output, what support are we providing to communication partners to empower self-determination, acknowledge intent, support consistency of message, and promote success?  When those around an AAC user have high expectations, are curious about how we feel, what we want and why, then we thrive because we feel valued. When we have knowledgeable communication partners who are patient and responsive, then we can relax, share our humour, make known our preferences and know we are not being tested every time we say something.  Knowledgeable communication partners don’t just happen; we need to find ways to support them.

The need for opportunities and affordances

We need others to be non-judgmental, whatever resources AAC users are using.  Don’t doubt what we are saying; instead, explore the topic with us. If it challenges your ‘norm for AAC users’ don’t doubt our competence, put us down or alternatively say we are an outlier, different to others.  There is nothing more harmful for any child than to be challenged, constantly tested, not being trusted and having opportunities reduced to match the expectations of others.

The need to presume competence

Please remember AAC users are working hard every day to do something that comes as second nature to anyone who speaks. Put ableism aside, instead presume competence and ask yourself how you would treat a speaking child or adult and do the same for us.

Let’s be ethical in our approach

So, should we, as AAC users, need to prove what we say is authentically our own words?  First, we shouldn’t have to, but if we do, then we need to be ethical about the way we do this.  The real measure should be the same as for anyone else, whatever method of communication they are using; agency is demonstrated through self-direction, intentional and consistent communication.

My own experiences

From a personal perspective, throughout life, I have often felt judged.  My mainstream secondary school thought my technology (20 years ago) gave me an unfair advantage over other students.  They tried to test me by making me answer a test paper by handwriting myself (an impossible physical task).  When I didn’t perform, they assumed my primary school had done my work previously,questioning if I had sat as a passenger.  They went as far as pronouncing I probably wouldn’t get any qualifications; how wrong they were. This pattern has repeated itself over time, even now, with post-graduate qualifications, people who don’t know me make assumptions.  This has to stop; we need to all embrace the technology available in everyday life and find ways to make it work to support and empower AAC users to learn and fulfil their own potential.

Please note: These are my words, although I used both the editor function in Word and Grammarly to check spelling and punctuation.  It was also uploaded to Co-Pilot (I’ve not used it before), who said my post was nuanced and drew together different perspectives. I did not take up any suggestions for improvement.