If people have the right tools in their hands, they may use them for creating unimagined results. Crowdsourcing is an excellent example. The web just provides a service and it is the combined effort of millions of users which creates the rich contents.

Other, perhaps more important, kind of tools are tools of thought. They help us to go beyond our ordinary thinking and thus allow us to discover unimagined results. Logic is one such tool. If mastered correctly, rules such as contraposition and reasoning by cases permit the detection of truths that are deeply hidden in the facts. It is in this way that Arthur Conan Doyle’s Sherlock Holmes amazed his friend and his readers.

However, many of our thoughts don’t revolve around what is true and what is false. Many of our thoughts revolve around what we like and what we don’t like. When revisiting the events of a day, a week, or a year, we check what we liked, we check whether we have missed some better choice and what we can do better next time. Did we choose the right car? Why didn’t we pick the great model that our neighbor has chosen? Did we miss something, forget some criterion when comparing the car models? Were we satisfied with the restaurant last night, the hotel chain, the vacation trip? Should we change our choices next time? So how should we compare those choices? What does matter for us? What are our preferences?

We compare options, analyze comparisons, extract preferences, revise them, consolidate them. We redefine what we mean by being better during those reflections. We become happy if we find clear preferences that justify our choices of the past and that will allow us to make choices in the future while taking into account the revisions that we decide during those reflections. However, if we don’t succeed to put our preferences in balance and to establish coherence between our past and future choices, the result may be frustration.

Preferences guarantee the consistency of an agent’s decision making behaviour in recurrent situations while taking the particularity of a situation into account. If the best option is not possible in a situation, a rational agent may, for example, consider the next best option until a best option has been identified among those that are possible in the given situation. Preferences thus guide an agent’s decision making behaviour and this even if the agent is not aware of the preferences as these may be encoded in form of rules or procedures. However, as soon as an agent needs to explain her actions or to revise her behaviour, it is necessary that the agent becomes aware of her preferences and represents them explicitly. Awareness of preferences is also necessary if a decision is taken collectively by group members. In that case, the agents communicate their preferences in the form of votes. As soon as preferences are represented explicitly, they can be analyzed, aggregated, consolidated, revised, and so on. In other terms, preferences become knowledge that can be manipulated as any other knowledge.  This knowledge does not describe the outside world and it is not true or false independent of the agent. Preference knowledge describes what the agent considers good and bad and it will permit us to explain and predict agent decisions if the agent is rational and not erratic.

Preferences can thus play different roles as explained in the article Prospects for Preferences by Jon Doyle published in Computational Intelligence 2004. The field of preference handling studies these roles and elaborates methods for the acquisition, representation, aggregation, consolidation, and revision of preference knowledge. It further studies methods for reasoning under preferences, for finding optimal choices under preferences, and for explaining the optimality of those choices under the given preferences. In this respect, preference handling elaborates tools of thought, which don’t seek to unveil hidden truths, but which seek a coherent understanding of an agent’s behaviour.