possible expenditure of whatever resource (time, money, energy . . . ) it takes. Efficiency with respect to relative goals is a matter of striking a balance between degree of achievement and expenditure. In the special case where the expenditure is fixed – say all the time available is going to be spent anyhow – efficiency con-sists in achieving the goal to the highest possible degree. Most discussions of information processing, whether in experimental psycho-logy or in artificial intelligence, have been concerned with the realisation of abso-lute goals. ‘Problem solving’ has become the paradigm of information processing. The problems considered have a fixed solution; the goal of the information-processing device is to find this solution; efficiency consists in finding it at the minimal cost. However, not all cognitive tasks fit this description; many tasks con-sist not in reaching an absolute goal, but in improving on an existing state of affairs. Hence, cognitive efficiency may have to be characterised differently for dif-ferent devices. Simpler information-processing devices, whether natural, such as a frog, or artificial, such as an electronic alarm system, process only very specific informa-tion: for example, metabolic changes and fly movements for frogs, noises and other vibrations for alarm systems. Their information-processing activity consists in mon-itoring changes in the values of a few variables. They could be informally described as engaged in answering a few set questions: ‘Is there a fly-like object within reach?’, ‘Is there a large body moving in the room?’ More complex information-processing devices, by contrast, can define and monitor new variables or formu-late and answer new questions. For the simpler devices, efficiency consists in answering their set questions at the minimal processing cost. Efficiency cannot be so easily defined for more com-plex devices such as human beings. For such devices, efficient information pro-cessing may involve formulating and trying to answer new questions despite the extra processing costs incurred. Formulating and answering specific questions must then be seen as subservient to a more general and abstract goal. It is in relation to this general goal that the efficiency of complex information-processing devices must be characterised. On the general goal of human cognition, we have nothing better to offer than rather trivial speculative remarks. However, these remarks have important and non-trivial consequences. It seems that human cognition is aimed at improving the individual’s knowledge of the world. This means adding more information, infor-mation that is more accurate, more easily retrievable, and more developed in areas of greater concern to the individual. Information processing is a permanent life-long task. An individual’s overall resources for information processing are, if not quite fixed, at least not very flexible. Thus, long-term cognitive efficiency consists in improving one’s knowledge of the world as much as possible given the avail-able resources. What, then, is short-term cognitive efficiency – efficiency, say, in the way your mind spends the next few seconds or milliseconds? This is a more concrete question, and one that is harder to answer. At every moment, many different cog-nitive tasks could be performed, and this for two reasons: first, human sensory

DOI link for possible expenditure of whatever resource (time, money, energy . . . ) it takes. Efficiency with respect to relative goals is a matter of striking a balance between degree of achievement and expenditure. In the special case where the expenditure is fixed – say all the time available is going to be spent anyhow – efficiency con-sists in achieving the goal to the highest possible degree. Most discussions of information processing, whether in experimental psycho-logy or in artificial intelligence, have been concerned with the realisation of abso-lute goals. ‘Problem solving’ has become the paradigm of information processing. The problems considered have a fixed solution; the goal of the information-processing device is to find this solution; efficiency consists in finding it at the minimal cost. However, not all cognitive tasks fit this description; many tasks con-sist not in reaching an absolute goal, but in improving on an existing state of affairs. Hence, cognitive efficiency may have to be characterised differently for dif-ferent devices. Simpler information-processing devices, whether natural, such as a frog, or artificial, such as an electronic alarm system, process only very specific informa-tion: for example, metabolic changes and fly movements for frogs, noises and other vibrations for alarm systems. Their information-processing activity consists in mon-itoring changes in the values of a few variables. They could be informally described as engaged in answering a few set questions: ‘Is there a fly-like object within reach?’, ‘Is there a large body moving in the room?’ More complex information-processing devices, by contrast, can define and monitor new variables or formu-late and answer new questions. For the simpler devices, efficiency consists in answering their set questions at the minimal processing cost. Efficiency cannot be so easily defined for more com-plex devices such as human beings. For such devices, efficient information pro-cessing may involve formulating and trying to answer new questions despite the extra processing costs incurred. Formulating and answering specific questions must then be seen as subservient to a more general and abstract goal. It is in relation to this general goal that the efficiency of complex information-processing devices must be characterised. On the general goal of human cognition, we have nothing better to offer than rather trivial speculative remarks. However, these remarks have important and non-trivial consequences. It seems that human cognition is aimed at improving the individual’s knowledge of the world. This means adding more information, infor-mation that is more accurate, more easily retrievable, and more developed in areas of greater concern to the individual. Information processing is a permanent life-long task. An individual’s overall resources for information processing are, if not quite fixed, at least not very flexible. Thus, long-term cognitive efficiency consists in improving one’s knowledge of the world as much as possible given the avail-able resources. What, then, is short-term cognitive efficiency – efficiency, say, in the way your mind spends the next few seconds or milliseconds? This is a more concrete question, and one that is harder to answer. At every moment, many different cog-nitive tasks could be performed, and this for two reasons: first, human sensory

possible expenditure of whatever resource (time, money, energy . . . ) it takes. Efficiency with respect to relative goals is a matter of striking a balance between degree of achievement and expenditure. In the special case where the expenditure is fixed – say all the time available is going to be spent anyhow – efficiency con-sists in achieving the goal to the highest possible degree. Most discussions of information processing, whether in experimental psycho-logy or in artificial intelligence, have been concerned with the realisation of abso-lute goals. ‘Problem solving’ has become the paradigm of information processing. The problems considered have a fixed solution; the goal of the information-processing device is to find this solution; efficiency consists in finding it at the minimal cost. However, not all cognitive tasks fit this description; many tasks con-sist not in reaching an absolute goal, but in improving on an existing state of affairs. Hence, cognitive efficiency may have to be characterised differently for dif-ferent devices. Simpler information-processing devices, whether natural, such as a frog, or artificial, such as an electronic alarm system, process only very specific informa-tion: for example, metabolic changes and fly movements for frogs, noises and other vibrations for alarm systems. Their information-processing activity consists in mon-itoring changes in the values of a few variables. They could be informally described as engaged in answering a few set questions: ‘Is there a fly-like object within reach?’, ‘Is there a large body moving in the room?’ More complex information-processing devices, by contrast, can define and monitor new variables or formu-late and answer new questions. For the simpler devices, efficiency consists in answering their set questions at the minimal processing cost. Efficiency cannot be so easily defined for more com-plex devices such as human beings. For such devices, efficient information pro-cessing may involve formulating and trying to answer new questions despite the extra processing costs incurred. Formulating and answering specific questions must then be seen as subservient to a more general and abstract goal. It is in relation to this general goal that the efficiency of complex information-processing devices must be characterised. On the general goal of human cognition, we have nothing better to offer than rather trivial speculative remarks. However, these remarks have important and non-trivial consequences. It seems that human cognition is aimed at improving the individual’s knowledge of the world. This means adding more information, infor-mation that is more accurate, more easily retrievable, and more developed in areas of greater concern to the individual. Information processing is a permanent life-long task. An individual’s overall resources for information processing are, if not quite fixed, at least not very flexible. Thus, long-term cognitive efficiency consists in improving one’s knowledge of the world as much as possible given the avail-able resources. What, then, is short-term cognitive efficiency – efficiency, say, in the way your mind spends the next few seconds or milliseconds? This is a more concrete question, and one that is harder to answer. At every moment, many different cog-nitive tasks could be performed, and this for two reasons: first, human sensory