Photo Courtesy: Image by Mohamed Hassan from Pixabay | For representational purpose only
Akshat Srivastava
First Year B Tech Student, Plaksha University
There has been a strange change in people’s behaviour recently: People are now talking to software in complete sentences. The previous rigid way of conversing is gone, with people no longer memorising commands or clicking through menus, but instead continuously talking, rephrasing, correcting, and improving their conversations to refine the software's responses. It may not seem like a big change at first glance, until you realise how it used to be just a few years ago. Software used to require people to conform to its logic. Now, artificial intelligence is designed to conform to people’s language, and this is not only a change in how software works but also in how people work with it.
Large language models are based on patterns of words and meaning; therefore, the words and meaning of a prompt can influence the output of the software. As an easily understandable example, two people may use the same AI service for the same purpose and yet achieve vastly different outcomes. Just like a conversation, the other side is the same, but how people choose to interact is not.
This is where literacy really comes into play beyond just being able to read and write–literacy is being able to express a thought or idea clearly, to ask a question, to comprehend an answer or response, and to recognize when words and/or sentences have fluency, yet no meaning. The basic idea of prompting is to improve your ability to work, as well as to refine your linguistic and cognitive abilities. Therefore, clear and well-structured prompts will produce the best results, and will also demonstrate intent and judgement.
What makes this even more interesting is that the change is not one-sided; rather, while humans are becoming more adept at working with and around machine responses to their prompts, humans are also becoming more inclined to interact with machines in such a manner as to mimic a conversation between humans. Interestingly, psychology has long illustrated that when information is easier to process, it is also more likely to feel convincing or familiar and satisfying. When language is fluent, direct and well-organised, it can influence the way in which a response is received before the essence of the response is judged or evaluated. Research into the phenomenon of processing fluency and the illusory truth effect points exactly in the same manner, suggesting that the ease of processing can influence judgements of credibility and confidence.
No longer do people necessarily rely on a search engine that offers links to sort through. Instead, there is now a summary of the search results near the top of the page, created by AI such as Gemini. There is a curated answer to the search query as the first response, and then the links. Google itself describes “AI Overviews” as a way to help people understand a topic quickly and ask follow-up questions within search. It’s easy to see why people might enjoy this format. This not only provides information but also a sense of closure and satisfaction. There is a sense of being understood by a conversational answer. With conversational AIs increasingly being used to assist people in a variety of tasks, the cognizance people need to have to prevent passive acceptance of answers is increasing day by day. Recent studies on trust and conversational agents suggest that the language of a conversational agent can influence trust, cognitive load, and competence. In general, language that is approachable, structured, and socially familiar can alter how an answer is perceived.
That is precisely why prompting should be taken seriously and not be dismissed as just another buzzword. The reality is that prompting is the place where language, psychology, and technology now meet. A good analogy for a prompt would be providing directions rather than a general indication of the location.
Prompting not only promotes better communication with machines, but also comes with the added benefit to effectively interpreting the machine. AI can sound coherent even when it's completely wrong, termed in the AI communities as “hallucinations”. A response may feel finished just because it’s fluent. In this case, the user’s scepticism becomes essential. If one considers prompting a literacy of some sort, then one must also consider this literacy to be one of scepticism. A user should be able to notice when an answer is avoiding the question, or in other words, when confidence is replacing substance.
Justifiably so, access to AI is not the same as proficiency in AI. What matters is the degree of mental engagement. A thoughtful user is not content with the first response. The prompt is refined, and the response checked. Information is sought to fill in the gaps, and assumptions are challenged. A slight variation in the prompt can lead to a completely different response. In all of this, the process of prompting is slowly transforming into an exercise in thinking rather than simply commanding.
The widespread acceptance and use of AI have changed our understanding of how we, as humans, can communicate. More importantly, they are “prompting” us to rethink how we view ourselves. If an answer appears accurate because it was given with an appropriate amount of fluency and at the appropriate moment, we are not focused on whether the information is true or false, but instead are relying on the convenience of accepting an answer without further consideration or evaluation. It could be argued that this is a more efficient way of thinking, but it certainly does not build a basis for thinking that would allow for a complete understanding of an answer.
Thus, this new form of literacy has the potential to provide us with higher-quality thinking, as opposed to simply obtaining more accurate responses from machines. Effective prompting is not only based on how well-developed a person's thinking process is, but also partially related to how much mental focus is directed toward creating an idea. Good prompting provides individuals with an opportunity to define, identify, and articulate what their meaning or intent is and what they may need to accept or hesitate to accept. Learning how to prompt well can become an automatic and useful habit that keeps your thinking active rather than passive.
That’s why this question is relevant to us now. If we become accustomed to talking to machines, then literacy can’t be merely the ability to read words on a page and write them back. It might also be the ability to initiate a conversation with a machine that speaks in a voice that is assured, friendly, and human, without giving up the very qualities that make human thinking valuable in the first place.
The real risk, then, is not only the possibility for error, but also the possibility for ease. The completeness of the response breeds the temptation to stop thinking when the response appears complete. As such, the initial phase of an emerging literacy is developing a habit (discipline) of taking a moment to stop, examine the initial response provided to you, look for missing information and think critically about how well the response either supports or replaces any thought processes that occurred before.
