|
Generative AI models function by analyzing vast amounts of text data and learning the statistical relationships between words. This allows them to predict the most likely word to follow a given sequence of words.
When you provide a prompt, you are essentially feeding the AI with a starting sequence. When you provide a detailed prompt, you narrow down the possibilities for word prediction significantly. Prompting is not about giving instructions to your Gen AI. It is more like a design skill thats helps you to shape how your AI thinks, creates and responds. |
Framework # 1: ECOSTAR method
This is a suggested framework. It is not THE way, but just A way. Also, all the components might not be needed all of the time.
The ECOSTAR framework is particularly useful for crafting prompts because it focuses on key elements that steer the AI towards generating outputs tailored for specific audiences and contexts. BUT ...please also note that this framework more than a tool for prompting. The (E)COSTAR framework closely aligns with core thinking skills too.
The ECOSTAR framework is particularly useful for crafting prompts because it focuses on key elements that steer the AI towards generating outputs tailored for specific audiences and contexts. BUT ...please also note that this framework more than a tool for prompting. The (E)COSTAR framework closely aligns with core thinking skills too.
COMPONENT |
DESCRIPTION |
Core thinking skill |
EXPERTISE |
Specify the domain of expertise to semantically access the LLMs focus. "You are an expert in political speech writing and rhetoric" |
- |
CONTEXT |
Establishing context involves presenting the background or rationale behind your inquiry. Instead of posing a standalone question such as “Write a proposal for investigating the correlation between height and earnings?”, you might begin with, “We are preparing a submission for an upcoming academic conference.” This contextual framing enables the AI to better understand the scope, purpose, and constraints of your task, thereby generating responses that are more relevant and aligned with your goals. |
Situational awareness & information gathering |
OBJECTIVE |
The objective articulates the specific outcome you wish to achieve. For instance, stating “Help draft an abstract for a conference paper on AI ethics” provides a clear directive that guides the AI’s response. A well-defined objective ensures that the output remains tailored to your academic needs. |
Goal setting & purpose-driven thinking |
STYLE |
Style refers to the preferred manner of expression in the AI’s response. Depending on your requirements, you may wish the AI to adopt the tone of a scholarly author, a grant writer, or a technical communicator. Specifying the style helps ensure that the language, structure, and rhetorical devices used are appropriate for the academic context and the expectations of your target audience. |
Perspective taking & tone awareness |
TONE |
Tone reflects the emotional or rhetorical character of the response. In academic settings, this often leans toward being formal, objective, and precise. However, depending on the nature of the conference or the intended audience, you might prefer a more persuasive, exploratory, or innovative tone. |
Analytical thinking & problem-solving |
AUDIENCE |
Identifying the audience involves clarifying who will be reading or evaluating the proposal. Whether the audience consists of domain experts, interdisciplinary scholars, or conference organisers, understanding their level of expertise and expectations allows the AI to adjust its language, depth, and framing. |
Empathy & audience awareness |
RESPONSE / REQUEST |
The response format specifies how you would like the information to be structured—such as a bullet-point outline, a formal abstract, or a structured proposal template. This is analogous to choosing the appropriate medium for academic communication, whether a journal article, a slide deck, or a submission form, and ensures the output is immediately usable within your workflow. |
Organizational thinking & presentation strategy |
Although ECOSTAR provides a strong foundation, you might need to iterate on your prompt a few times to achieve the exact level of output you desire. The initial attempts might be more open-ended than you envisioned. By understanding how generative AI models function as statistical predictors and how prompts act as stepping stones, you can unlock greater creative potential from these tools
Framework # 2: TCREI Method
TCREI stands for: Task, Context, Resources, Evaluate, Iterate
Description |
||
T |
TASK |
Define exactly what you want Gen AI to do. eg. Create a rubric for an essay assignment that measures the following criteria .... |
C |
CONTEXT |
Feed the background information. Your organisation details, target audience, brand voice, launch date. More context = better results. |
R |
RESOURCES |
Provide examples of what good looks like. Screenshots, competitor posts, previous content that worked. Show, don’t just tell. |
E |
EVALUATE |
Read the output like a skeptical teacher/ student Does it hit the mark? What’s missing? What’s wrong? |
I |
ITERATE |
Refine based on your evaluation. “Make the tone more conversational” or “add specific pricing details.” |
Exemplified as:
T – TASK: What do you want Gen AI to do?
“Write a persuasive 500-word campaign speech in English for Magdalena Andersson, aimed at rural voters in Jämtland and Västerbotten.”
Further instruction:
C – CONTEXT: What background should the AI know?
Magdalena Andersson is delivering a summer speech tour as opposition leader. Rural voters feel disconnected from Stockholm and are concerned about:
R – RESOURCES: What are examples of what “good” looks like?
Provide inputs such as:
Tone example: “Use language similar to Obama’s rural town hall speeches—warm, vivid, grounded in real-life challenges.”
Style example: “Refer to the metaphor of navigating a whiteout on skis to describe unpredictable political changes.”
Structure reference: Opening hook → 2 stories → rallying conclusion.
Cultural touchstones: Taco Fridays in Sweden, weather unpredictability in the mountains.
E – EVALUATE: Read the AI output critically.
Ask:
Does the speech reflect the values of fairness, tradition, and local pride?
Are the metaphors and anecdotes meaningful and not forced?
Is the tone sincere, slightly humorous, and never patronising?
Does it sound like a human political leader—not like a press release?
Example Evaluation Feedback:
"The taco anecdote feels generic—localise it more to northern Sweden. The final paragraph could be more emotionally resonant."
I – ITERATE
Refine the output based on evaluation. For example:
T – TASK: What do you want Gen AI to do?
“Write a persuasive 500-word campaign speech in English for Magdalena Andersson, aimed at rural voters in Jämtland and Västerbotten.”
Further instruction:
- Use metaphor-rich language (e.g. weather, mountains, skiing).
- Include two anecdotes: one about an unexpected mountain weather event (teamwork) and one about tacos (humour, inclusivity).
- End with a strong, shareable message about hope, resilience, and solidarity.
C – CONTEXT: What background should the AI know?
Magdalena Andersson is delivering a summer speech tour as opposition leader. Rural voters feel disconnected from Stockholm and are concerned about:
- Public service access
- Economic insecurity
- Societal cohesion
R – RESOURCES: What are examples of what “good” looks like?
Provide inputs such as:
Tone example: “Use language similar to Obama’s rural town hall speeches—warm, vivid, grounded in real-life challenges.”
Style example: “Refer to the metaphor of navigating a whiteout on skis to describe unpredictable political changes.”
Structure reference: Opening hook → 2 stories → rallying conclusion.
Cultural touchstones: Taco Fridays in Sweden, weather unpredictability in the mountains.
E – EVALUATE: Read the AI output critically.
Ask:
Does the speech reflect the values of fairness, tradition, and local pride?
Are the metaphors and anecdotes meaningful and not forced?
Is the tone sincere, slightly humorous, and never patronising?
Does it sound like a human political leader—not like a press release?
Example Evaluation Feedback:
"The taco anecdote feels generic—localise it more to northern Sweden. The final paragraph could be more emotionally resonant."
I – ITERATE
Refine the output based on evaluation. For example:
- “Add a reference to a known mountain in Jämtland or Västerbotten.”
- “Make the taco story more culturally specific—mention a summer festival or youth group dinner.”
- “Tighten the final sentence to make it more shareable.”