Broadly speaking we use the term 'model' to refer to any structured knowledge that accurately reflects – and helps us to make sense of – the world. Models exist both internally as 'mental models' and externally as 'cognitive artifacts'. Cognitive artifacts can take many forms: written texts, spoken stories, graphs, diagrams, pictures, videos, spreadsheets, equations, computer-simulations, etc. While these different kinds of models vary greatly in their form and function, they all share certain desirable properties.

A good model should be:

• Salient (1): Since no model can represent everything, it must selectively represent those things most relevant to the task at hand.
  
  
• Accurate(2,6): The model should precisely encode the actual state of affairs and not an erroneous or biased view.
  
  
• Complete yet Parsimonious (2): The model should be as simple as possible, but no simpler. It should concisely capture all the relevant dimensions of the problem without squeezing out the opportunity for serendipitous or creative insight.
  
  
• Perceptible (4): Models should be appropriately displayed in high fidelity as they won't be much use if we can't clearly see, hear, or feel them.
  
  
• Understandable (5): Once we perceive the model we must be able to make sense of it; it shouldn't be too complicated or unfamiliar for us to understand.
  
  
• Predictive(8): The model should not only depict what is, but should also predict what will happen; or at least give some insight into future possibilities.
  
  
• Falsifiable (9): The model should be formulated such that it is possible, through experimentation, to both confirm and disconfirm the model's accuracy and predictive power.
  
  
• Emotive (7): In addition, the model may convey a subjective feel for the emotional and value-laden connotations of the situation being modeled.
  
• Inspiring (7): Because people are drawn to and inspired by thoughtful design, models should be elegant, i.e. they should synergistically combine style and substance.
  
  
• Memorable (5): Models are not of much use if they pass quickly from the mind, or if they cannot be used as a mnemonic device. Models should be easily accessible for future reference and to refresh our understanding.
  
  
• Flexible (3,9): As all models are, to some degree, inaccurate, irrelevant, mistaken, time-sensitive etc., they should be open to recursive revision to reflect new data, our growing understanding, or our evolving needs.
  
  
• Coherent (3): Models do not exist in isolation but in interlocking systems, thus any particular model should be coherent with other related models.
  
  
• Productive (10): Ultimately, the model has a purpose: the production of effective action. A good model should help define our goals and then specify the actions necessary to reach them.
  
  
• Useful (1-10): Usefulness is the sum of the above properties and the degree to which they combine to promote understanding and effective action. It is important to note that the most accurate, or the most complete, or the most elegant model is not necessarily the most useful. All models are incomplete. All models a compromise. The model maker's art lies in making those shrewd trade-offs that will render the model most useful to the problem at hand.