Introduction First of all, I will connect to my article “Decision making”, published in this Journal some years ago1. In that issue I pointed out some aspects, among them focalization, pseudodiagnosticity and framing-effect, which, according to some research works, significantly influence decision making. Focalization, and pseudodiagnosticity are two inter-related ways of information processing: the first one concerns the tendency to consider only some elements of a situation and to leave out all the others; the second one consists of leaving out important aspects for a diagnosis (diagnosis widely considered, beyond the medical sphere). Focaliza- tion and pseudodiagnosticity imply that, in deci- sion making, the subjects are satisfied with an in- complete research of the possible alternatives2. Fo- calization and pseudodiagnosticity are explained by “Mental Models Theory” (MMT), stating that “…we are unconsciously satisfied with simplified models of reality and … such models guide our thinking and behaviour”2. As for framing-effect, it is the conditioning of the decision by the verbal formulation modalities of the alternatives. First data of a research on focalization, pseudodiagnosticity and framing-effect I will explain now the concepts of focalization, pseudodiagnosticity and framing-effect by the pro- visional data of a research work on decision mak- ing I am carrying out on undergraduate students. This research requires a yes/no answer, with possi- ble addition of observations, to the following four questions: • Question No. 1 In your city, in the last month, an Indian holy man healed 15 people. A famous physician, who is in- terested in the same pathology, healed 4 people. Would you recommend an examination by the Indian holy man to a friend of yours, who suffers from the problems treated by both the holy man and the physician? • Question No. 2 Imagine that you have just graduated and you receive a job offer from an important multinational compa- ny. You are offered an interesting job and a € 5,000 initial monthly salary. You are also required to make an immediate decision. Would you accept the offer? • Question No. 3A During a party, in a newlywed’s couple’s home, you meet some nice boys and girls, who invite you to ride their car and spend the rest of the night in a club in an area you like very much. Would you accept? • Question No. 3B During a family party you make friends with some boys and girls, who invite you to ride their car and spend the rest of the night in a club in an area you like very much. Would you accept? The cognitive error in decision making Pier Luigi Baldi Full professor of General Psychology, Catholic University of Milan This issue deals with the partial data of a research in progress on focalization, pseudodiagnosticity and fram- ing-effect in decision making, followed by the most im- portant results of some experiments about the emotional aspects of the choice, and ends by stressing the potential contribution of the artificial neural networks to the medi- cal diagnosis. ABSTRACT emergency care journal organizzazione e formazione em er ge nc y ca re jo ur na l - o rg an iz za zi o ne , c lin ic a, r ic er ca • A nn o V I n um er o 2 • G iu gn o 2 01 0 • w w w .e cj .it Materiale protetto da copyright. Non fotocopiare o distribuire elettronicamente senza l’autorizzazione scritta dell’editore. 13 Questions 3A and 3B are parallel and convey the same information, but there are some formal dif- ferences among them, aiming to elicit more confi- dence in reading question 3A than in reading ques- tion 3B (e.g., “the newlywed’s couple’s home” of question 3A becomes “family” in question 3B) and so promoting the framing-effect. Question 3A was put to a subgroup of subjects, question 3B to the other subgroup. Table1 shows the results obtained so far. In my opinion, the most interesting information in Table 1 concern the answers to question No. 2 and the comparison between the answers to question No. 3A and the ones to question No. 3B. In regards to the data of question No. 2, they are probably referable to the focalization on the in- teraction between the promised high pay and the prospect of an interesting job; nearly all the boys accept the job offer without mentioning, in the comments’ space, what the demands of the mul- tinational company may be (e.g. one can wonder about the job site location and how frequently homecoming is possible, what are the risks in- volved in the job even if it’s interesting, etc.) and show a clear pseudodiagnosticity. The comparison between answers 3A and 3B can be interpreted in the light of the framing-effect: the few formal changes in question 3B have been enough to obtain the overturning of the percent- ages of answers. Finally, I think we can easily admit that focaliza- tion and pseudodiagnosticity explain the not insig- nificant percentage of subjects who answered “yes” to question No. 1. These ones didn’t realize that the data are incomplete, since the number of peo- ple who respectively consulted the holy man and the physician isn’t stated: the holy man could have by chance healed 15 people over dozens of them, while the “famous” physician, particularly engaged in scientific activities, could have healed 4 over 5 patients examined in the last month. If the focalization and the pseudodiagnosticity may be largely ascribed to a simplification process, that can be linked to people’s modest cognitive re- sources, process that allows to deal with otherwise unmanageable problems and decisional situations3 (Simon, 1982), non-cognitive variables are crucial on other mistakes, which we make over decision making. Emotions and decision Several experimental works emphasized the fre- quent indecision and inconsistency of people’s preferences4, preferences influenced by personality aspects, above all by emotional aspects5. The results of many psychological researches, be- ginning from Zajonic’s early researches of the second half of last century up to more recent works, point- ed out the inadequacy of an explanation of many decisions exclusively based on rational founda- tions and the importance of the approach to the de- cision making, considering the emotional elements. Some studies on people suffering from anxiety dis- TAB. 1 Answers to the questions (%). Questions No. 1 No. 2* No. 3A No. 3B Answers Yes No 32 68 84 12 36 64 62 38 * No answer: 4%. TAB. 2 Equidistribution test of the yes/no answers to questions 3A and 3B. 3A 3B Tot. χ2 Yes No 4 7 9 5 13 12 6.09* Tot. 11 14 25 No. = 25; Chi-square test with Yates’ correction for continuity; Df = 1; *p < 0.05. organizzazione e formazione em er ge nc y ca re jo ur na l - o rg an iz za zi o ne , c lin ic a, r ic er ca • A nn o V I n um er o 2 • G iu gn o 2 01 0 • w w w .e cj .it Materiale protetto da copyright. Non fotocopiare o distribuire elettronicamente senza l’autorizzazione scritta dell’editore. 14 orders showed their outstanding tendency to con- sider that adverse events are very likely to happen and to perceive themselves in a risk situation6,7. Affective states experimentally caused in non anx- ious people also allowed to observe that, as a con- sequence of unpleasant emotional states, the per- ception of the risk rises, causing a significant spin- off on the performances and on the consideration of one’s choices8,9. According to Bower & Cohen10, affective states may act as a filter and the prevailing emotional state causes only certain aspects of the decisional context, and no others, to be selected. Possible consequences are decisions highly conditioned by the prevailing momentary emotion. In an experimental work, Wright & Bower11 caused subjects of the experimental group to be in a pleas- ant emotional state, by asking them to recall a situ- ation in which they had been particularly happy, or an unpleasant emotional state, by asking them to recall a particularly unhappy situation. Afterwards the same subjects, in hypnotic state, had to consider the probability a series of events to happen, where in half cases they were protagonists; the results of the events would have been positive or negative. In comparison with the control group, without any caused emotional state, the subjects who had recalled happy situations tended to over- estimate the probability of positive events and to underestimate the risk of negative events, whereas the subjects who had recalled unhappy situations behaved in the opposite way. The neurosciences also emphasized the affective aspects of decision making. E.g., Damasio12 ob- served the impossibility of rational decisions, in spite of the completeness of the intellectual facul- ties, in patients showing lesions of the frontal cor- tex, unable to set links between the effects of their action and the related emotions. On the basis of the obtained outcomes, Damasio12 states that the emotional experience guides the de- cision. Certainly, in the next few years the technological strategies will be improved, to support decision making, strategies which the cognitive science gave a significant contribution to. This area includes re- searches about the neural networks. Neural networks and decision making processes Officially, the “Cognitive Science” was born in 1977, year in which the journal Cognitive Science was founded; its two paradigms were modular- ism and connectionism. Connectionism had more followers, while Fodor’s modularism, by hypoth- esizing a modular cognitive architecture, above all claimed neuropsychologists’attention. The connectionist approach imposed itself also be- cause of the substantial inadequacy of the previous computational models to simulate the functioning of the human mind; it is characterized by “brain style” models which are artificial “neural net- works”; these networks simulate biological neural networks and, so, the nervous system. In fact, the computer can carry out tasks which are very complex for the individual, as mathematic calculi, solution of logical problems, memorization of lots of data, but it cannot reproduce the tasks that our mind perform very easily on a daily ba- sis as recognizing one object among others from few characteristics, recognizing words written by different handwritings or upturned letters, making decisions based on few information. In short, com- puters, unlike the human mind, cannot work with imprecise or incomplete data. Some important differences between the nervous system and a serial system of information process- ing are as follows13: • Parallel processing of the information, while the traditional computers processing of each datum is individual and serial; that explains a greater velocity of the brain in executing tasks, as visual recognition of the objects, which re- quires simultaneous processing of a lot of data. • In the nervous system many neurons execute the same task; the processing is therefore dis- tributed among many elements, as the intracel- lular registration or cerebral activity visualisa- tion techniques show. In addition, each neuron can participate in different types of elaboration both simultaneously and serially. • Unlike computers, which access to each mem- orized information through a numerical ad- dress, the human brain accesses to its memory through the content of the information. • The nervous system performs tasks without be- ing programmed and automatically learns from experience or with an expert teacher’s help. Although the artificial neural networks operate quite differently from the serial computers, it’s pos- sible to simulate an artificial neural network on any type of computer. An Artificial Neural Network (ANN) is a model which simulates the biological neural networks. It is composed of aggregations (or nets) of simple ele- ments, named artificial neurons or unities or knots organizzazione e formazione em er ge nc y ca re jo ur na l - o rg an iz za zi o ne , c lin ic a, r ic er ca • A nn o V I n um er o 2 • G iu gn o 2 01 0 • w w w .e cj .it Materiale protetto da copyright. Non fotocopiare o distribuire elettronicamente senza l’autorizzazione scritta dell’editore. 15 ta rg et 2 /1 0 Negli ultimi 40.000 anni i nostri geni sono cambiati poco. Erano stati disegnati per adattare l’uomo a vivere nelle caverne, a cibarsi di prodotti vegetali presenti in natura e a mangiare la carne solo quando riusciva a cacciare. Oggi invece la vita è cambiata moltissimo e il progresso è riuscito ad agevolare in maniera determinante le attività dell’uomo, facen- do però insorgere malattie in passato sconosciute. Infatti la riduzione dell’attività fisica e l’aumentato ap- porto di alimenti ricchi di calorie e poveri di fibre hanno favorito l’insorgenza di malattie come l’obesità, l’arteriosclerosi, l’ipertensione arteriosa e il diabete. Senza ombra di dubbio una corretta alimentazione nella prima infanzia riesce a prevenire gran parte delle malattie. È importante che le persone, soprattutto coloro che hanno il compi- to di educare i bambini (genitori, nonni, insegnanti), abbiano ben chiari i concetti sulla nutrizione umana e quali siano i com- portamenti alimentari da seguire. Il volume vuole essere un concreto aiuto per tutti coloro che voglio- no capire meglio quanto sia importante il fattore alimenta- re, per vivere bene e prevenire eventuali malattie. Gli alimenti e la loro origine • Gli alimenti sono il combustibile bio- logico che fa funzionare la nostra macchina • Gli elementi chimici che costituiscono gli alimenti e il metabolismo • Le chilocalorie • La digestione, l’assorbimento, il catabolismo e l’anabolismo • Le carat- teristiche degli alimenti e la dieta mediterranea • Tutti gli elementi chimici che costituiscono gli alimenti • Determinazione della razione alimentare giornaliera o dieta • Come vedere se il nostro peso è nella norma • La dieta normale • Il sovrappeso e l’obesità • Diete speciali • Come ritardare l’invecchiamento • Danni provocati al no- stro organismo dall’uso eccessivo di carboidrati e di sale • Attività fisica sportiva e ginnastica • Alimenti naturali e diete macrobiotiche • La conservazione degli alimenti • Le alterazioni degli alimenti e gli agenti chimici, fisici e biologici che possono determinarle • Le varie tecniche di conservazione degli alimenti • Nozioni circa l‘uso degli alimenti conservati • Come cuocere e preparare gli alimenti • I diversi modi di cuocere i cibi • Appendici Giuseppe Murabito Medico chirurgo Specialista in Igiene e in Odontoiatria Saper mangiare per vivere e invecchiare bene Principi basilari per una corretta alimentazione Piano dell’Opera SCHEDA TECNICA 17 x 24 cm • 272 pagine ISBN: 978-88-7110-267-2 Prezzo di listino € 25,00 Novità 2010 La medicina d’emergenza e urgenza è una nuova specialità che si affaccia alla ribalta nel mondo sempre più diversificato dell’assistenza. È necessario avere medici d’emergenza e urgenza sempre più preparati, che si possano confrontare alla pari con i colleghi specialisti cardiologi, rianimatori, chirurghi o traumatologi. Le difficoltà dello specialista in emergenza e urgenza sono tante, perché di fondo si confronta con tutte le patologie acu- te e quindi mantenere una competenza ed una conoscenza approfondita sulle tematiche delle emergenze richiede tem- po, passione e studio. Passare da un possibile caso di addome acuto ad un dolore toracico è per il medico una quotidianità e richie- de una straordinaria capacità di focalizzare diagnosi e terapie. Sapendo che il rischio di una mancata diagnosi può essere fatale per il paziente e conoscendo quali sono gli errori più comuni dei medici d’urgenza, gli autori hanno pensato di raccogliere una serie di score che possano aiutare la valutazione clinica, sia per sostenere la pro- babilità diagnostica che per stratificare la prognosi. La maggior parte degli score riportati sono validati nella letteratura da ampi studi clinici che consentono la scelta della dimissione piutto- sto che del ricovero e della relativa strategia terapeutica. Questo tascabile è un sintetico, concreto e valido aiuto per tutti i medici che operano nel settore. Tiziano Lenzi UOC Pronto Soccorso Medicina D’Urgenza Dipartimento di Emergenza AUSL Imola Andrea Tampieri UOC Pronto Soccorso Medicina D’Urgenza Dipartimento di Emergenza AUSL Imola Maria Chiara Cantarini UOC Pronto Soccorso Medicina D’Urgenza Dipartimento di Emergenza AUSL Imola Scores Collana “Decidere in Medicina” Tiziano Lenzi Andrea Tampieri M. Chiara Cantarini Tiziano Lenzi - A ndrea Tam pieri - M . C hiara C antarini Scores C linici in M edicina d’U rgenza L 10,00 Scores Clinici in Medicina d’Urgenza Collana di Medicina d’Urgenza a cura di Ivo Casagranda Dal sapere al saper fare SCHEDA TECNICA 14 x 18 cm • 64 pagine ISBN: 978-88-7110-266-5 Prezzo di listino € 10,00 Piano dell’Opera Patologia gastroenterica - Patologia cardiovascolare - Eventi cerebro- vascolari - Scores predittivi di embolia polmonare - Polmoniti - SIRS e sepsi - Traumatologia - Bibliografia Saper mangiare per vivere e invecchiare bene Clinici in Medicina d’Urgenza Via Candido Viberti, 7 - 10141 Torino Novità 2010 or processors13; the transmission of the signal from one knot to another one is modulated by synapses, which can amplify the signal or reduce it. In a neural network there are at least two types of unities: the input units and the output units. Nev- ertheless multilayer neural networks are required for complicated tasks. The intermediate knots are called hidden units, since they are not in contact with the external and don’t produce directly a re- sponse. The various types of neural networks may be sim- ply divided into supervised and unsupervised net- works. The first ones, unlike the second ones, are trained to respond to the inputs, to be processed by outputs which lean gradually to the expected outputs. A particularly used training method is the error back-propagation algorithm*, by which the weights of the connections among the knots of the network are progressively modified. As for the unsupervised networks, they are based on training algorithms, which modify the weights of the network by using only the input data. Among them, Kohonen’s networks (SOM: Self-Or- ganizing Maps), are well-known to the specialists in the field and feature many applications. E.g., a SOM can map the input data and cluster them by a certain criterion. The following is a simple example, derived from Johson Laird (1993) and adapted by Anolli & Legrenzi (op. cit.). I think that it can give some useful explanation of the connectionist networks (Figures 2 and 3). The network represented in Figure 2 symbolizes an inclusive disjunction (A or B or both); suppose that two input units, A (mother) and B (nanny), produce an affective stimulation, which reaches force 1; the child gives a response as output (e.g., blowing a kiss) if he/she is encouraged by his/her mother or by his/her nanny or by both of them, as A and B’s encouragement, and even more so A and B’s joint encouragement exceeds 0.5 threshold value. The child doesn’t blow a kiss only if he/she is not stimulated. The network represented in Figure 3 symbolizes an exclusive disjunction (A or B, not both); in this case we hypothesize a man, who kisses their spouse or their lover, not both at the same time. Inhibitory Fig. 3 - Representation of an exclusive disjunction. * The word “algorithm” derives from the name of the Persian math- ematician Al-Khuwarizmi (9th century); it indicates a systematic problem solving procedure by a finite number of steps. E.g., the softwares are algorithms consisting of logical and algebraic opera- tions, written in a language which can be understood by computers. Fig. 1 - An example of a tree layers neural network. Input layer Hidden layer Output layer Fig. 2 - Representation of an inclusive disjunction. Output unit (kiss) Input unit A (mother) Input unit B (nanny) Output unit (kiss): 0.5 threshold value +1 +1 Input unit B (lover) Input unit A (spouse) Hidden unit: 1.5 threshold value +1 +1 +1+1 -2 0.5 0.5 1.5 organizzazione e formazione em er ge nc y ca re jo ur na l - o rg an iz za zi o ne , c lin ic a, r ic er ca • A nn o V I n um er o 2 • G iu gn o 2 01 0 • w w w .e cj .it Materiale protetto da copyright. Non fotocopiare o distribuire elettronicamente senza l’autorizzazione scritta dell’editore. 18 hidden unit, whose threshold value is 1.5 explains this condition: if the man is in front of their spouse or their lover (inhibitory hidden unit will not put itself in action, for its threshold value is 1.5) will express themselves with a kiss; if the same man is in front of both of them, inhibitory hidden unit will put itself in action, for it will receive a stimulation whose value is 2, that is more than its threshold value, and will send to the output unit a stimulation equal and opposite to the sum of the activations from A and from B. That situation will inhibit the response of the output unit and the man’s kiss. Granted that medical diagnosis is a complex proc- ess, the artificial neural networks, according to the experts of this area, can be a great help for both the diagnosis of the pathologies and the subsequent prognosis. A neural network can operate on images (X-ray plates), as well as on symbolic data, which represent symptoms and data13. It’s well known the Anderson’s so-called “instant physician”**, that is a neural network for diagnosis and patients’ treatment. I think that the chance to produce a “neural net- work-artificial physician” in future cannot be “a priori” excluded; such a network could gather the experience of the best physicians in the world counterbalancing the fact that a physician who quits his/her activity deprives the community from his/her experience. However it goes, it’s presumable that an “artificial opinion” may be helpful to the clinician’s decision process. ** Attributed by Hecht-Nielsen16. References 1. Baldi PL La decisione. Decision making. Emerg Care J; 2006; 3: 18-21. 2. Anolli L, Legrenzi P. Psicologia generale. Il Mulino, Bologna, 2003. 3. Simon HA. Models of bounded rationality. MIT Press, Cambridge (MA), 1982. 4. March JG. Exploration and exploitation in organizational learn- ing. Organization Science 1991; 10: 71-87. 5. Kahneman DE, Ritov I, Schkade D. Economic preferences or attitude expressions? An analysis of dollar responses to public issues. J Risk Uncertain 1999; 19: 203-235. 6. Engelhard IM, Van den Hout MA, Arntz A, McNally RJ. A longi- tudinal study of ‘intrusion-based reasoning’ and post-traumatic stress disorder after exposure to a train disaster. Behav Res Ther 2002; 40 (12): 1415-1424. 7. Engelhard IM, Macklin M, McNally RJ et al. A. Emotion and ‘intrusion-based reasoning’ in Vietnam veterans with and with- out chronic post-traumatic stress disorder. Behav Res Ther 2003; 39(11): 1339-1348. 8. Gasper K, Clore GL. The persistent use of negative affect by anxious individuals to estimate risk. J Pers Soc Psychol 1998; 74: 1350-1363. 9. Scott WD, Cervone D. The impact of negative affect on per- formance standards: evidence for an affect-as-information mechanism. Cogn Ther Res 2002; 26: 19-37. 10. Bower GH, Cohen PR. Emotional influences in memory and think- ing: data and theory. In: Clark MS, Fiske ST (eds). Affect and Cognition: The 17th Annual Carnegie Symposium on Cognition. Erlbaum, Hillsdale (NJ), 1982. 11. Wright W, Bower GH. Mood effects on subjective probability as- sessment. Organ Behav Hum Decis Process 1992; 52(2): 276-291. 12. Damasio AR. Descartes’ Error: Emotion, Reason and the Human Brain. Avon, New York, 1994. 13. Floreano D. Manuale sulle reti neurali. Il Mulino, Bologna, 1996. 14. Johnson-Laird PN. The computer and the mind: an introduction to cognitive science, 2nd ed. William Collins Sons & Co., London, 1993. Trad.it. La mente e il computer. Introduzione alla scienza cognitiva, 2a ed. Il Mulino, Bologna, 1997. 15. Anderson JA. Cognitive Capabilities of a Parallel System. In: Bi- enenstock E, Fogelman-Souli F, Weisbuch G (eds). Disordered Systems and Biological Organization. Springer-Verlag, Berlin, 1986, NATO-ASI Series. 16. Hecht-Nielsen R. Neurocomputing. Addison-Wesley, Reading (MA), 1990, p. 354. organizzazione e formazione em er ge nc y ca re jo ur na l - o rg an iz za zi o ne , c lin ic a, r ic er ca • A nn o V I n um er o 2 • G iu gn o 2 01 0 • w w w .e cj .it Materiale protetto da copyright. Non fotocopiare o distribuire elettronicamente senza l’autorizzazione scritta dell’editore. 19