The AI revolution is rapidly transforming the world, with prompts playing a crucial role in this change. A prompt is a snippet of text, question, command, or information that tells an AI model what response or action is desired. It serves as the starting point for the AI’s generation process, directing it to produce content relevant to the input provided.
For a while, “prompt engineering” was a major buzzword. However, branding this as “engineering” is somewhat misleading. While it shares characteristics with traditional engineering, such as problem-solving and optimization, it lacks formal structure and standardization.
The term “prompt engineering” benefits various stakeholders by providing professional recognition and legitimacy. It differentiates the role from casual uses of AI prompts and suggests a systematic approach to creating and optimizing them. This can be appealing to businesses integrating AI solutions and supports the development of specialized educational programs. Individuals with skills in AI and natural language processing can leverage the “engineering” label to command higher salaries and better job opportunities.
Despite its name, prompt engineering lacks a formal methodology. It involves experimenting with the language model to see how it responds. If it returns a consistent and interesting response, a reliable prompt is developed. This process is about trial and error, hoping to achieve a useful and replicable result.
Initially, prompt engineering took a formal turn with the development of templates and variables to simulate a formal language. However, this approach contradicted the rationale behind developing intelligence models. Overexposure to engineered prompts can lead to “contamination” or bias, resulting in non-human and incomprehensible language.
Despite the initial hype, the trend of prompt engineering declined, and the Socratic dialogue model of questions and answers prevailed. This shift has positive potential. The need to conduct real dialogues with AI, being precise and specific but not formulaic, may in turn improve human-to-human communication, enhancing our conversational and rational abilities. This could mitigate the polarizing language encouraged by social media algorithms, encouraging more thoughtful and nuanced discussions.
Prompt engineering and social media
In 2020, Twitter (now X) tested a new feature that prompted users to reconsider potentially offensive tweets before sending them. Users had the chance to edit, delete, or send the reply as is. The tests found that 34% of people revised their initial reply or decided not to send it at all. After being prompted once, people composed 11% fewer offensive replies in the future.
Twitter also temporarily changed the retweet function ahead of the US elections, encouraging users to add their own commentary before sharing. This made the retweet button default to a “quote tweet” option, aiming to increase the chances that users would contribute their own thoughts, reactions, and perspectives to the conversation.
The current status of this experiment is unclear, but it highlights the future potential of prompts to influence behavior positively. For now, we are living in the age of the hashtag and not of the cool-off prompt. Hashtags – more so than prompts – act as behavioral commands rather than as tools for communication. They make us intellectually lazy, flatten thought, oversimplify reality, insist that we pick a side, adopt a label and share it, leaving no room for complexity: #BLM (Black Lives Matter) or #AllLivesMatter, #TransRightsAreHumanRights or #SaveWomensSports, #ClimateCrisis or #ClimateHoax – and #StandWithIsrael or #FreePalestine.
The writer is senior lecturer of media studies at the College of Management Academic Studies in Rishon Lezion. She is the author of Context Blindness: Digital Technology and the Next Stage of Human Evolution.