Causal Judgement and Blocking Experiment

Causal Judgement and Blocking Experiment Michael Arruzza Abstract Recent results of studies which aim to gain a greater understanding of how blocking works and the mechanisms involved have led to some conflicting results. Some research suggest that blocking may be the result of assumptions about the additive nature of outcomes. Others think that it may have more to do with associative learning. This experiment investigates blocking in causal learning with human participants with the aim to addresses two main questions. Is blocking affected by encouraging or discouraging assumptions about the additive nature of the outcome? How do the results differ if we separately measure the memory for specific cue-outcome pairings? Results of the study show that there is evidence for both ideas and they may both contribute to the phenomenon of blocking. Introduction A model of the blocking effect was first developed by Kamin in his work with conditioning animals (Kamin, 1969). Kamin identified that training with one cue considerably reduced performance to another cue that was added to the first cue, in a second phase of training. A typical example of blocking would go something like this. Cue A is paired with an effect (A+). In a second phase a compound of two cues, the former cue A and new cue B is paired with the effect (AB+). Decisions on the strength of cue B are effected or â€Å"blocked† by information about the causal relationship of cue A. This has been described as the blocking effect. He theorised that learning may be driven by surprise, if a cue did not provide new information about the contingency between events and the environment it was ignored. Others researchers more recently such as Lovibond (Lovibond P, 2003) have tried to explain blocking in terms of inferential reasoning based upon a relatively simple set of propositions where the observer can deduce though logic that the blocked cue is not a cause of the outcome. Using an allergic reaction to foods as an example this would occur if someone assumed that the patients allergies are additive. When two foods are eaten which cause an allergic reaction they should cause a greater reaction then if only one of the two foods caused a reaction. If one does not observe an increase in the severity of the outcome when B is eaten at the same time as the allergenic food A, then one can deduce that B does not contribute to the allergic reaction. For example if eating a banana causes an allergic reaction rated as 5 and eating a banana and an apple causes a reaction of 5, then the apple has not made the reaction worse and is not the cause of the reaction itself. Lovibond (Lovibond P, 2003) and his colleges also observed that pre-training with explicit instructions that encourage outcome additivity assumption enhance the blocking effect. It is possible that the blocking effect is product of deductive reasoning and it sometimes unreliable response might be due to some participants assuming outcome aditivity while others do not. However a statistically robust blocking effect has still been observed even after explicit non-additive pre-training (Mitchell Chris, 2006). But his blocking effect is significantly smaller than after additive pre-training. In this study done by Mitchell et all multiple blocked use and control cues were implemented. The study only used a non-additive pre-training group and did not include an additive group. This would have been beneficial as it would have been great to contrast the two groups. After non additive pre-training participants were worse off at recalling the outcome associated with the blocked cue. Mitchell goes on to suggest from the data that the blocking effect that has previously been observed after non-additive pre-training may be the result of a non-rational process such as a failure to retrieve the outcome associated with the blocked cue. The ‘blocking affect’ may results from a failure to encode the B–outcome relationship during training. This concept tried to explain blocking as an associative learning theory. This research aims to better understand the blocking phenomenon. In particular to reconcile the two somewhat contradicting explanations outlined above that either blocking is the result of deductive reasoning or it is a part of a failure to encode cue-outcome associations. Is it the product of one of these two ways of thinking about blocking, can they co-exist or are they mutually exclusive? Based on previous research: It is hypothesised that both additive and non-additive groups would show reliable blocking in memory for both groups It is also hypothesised that there will be a weaker causal attribution for blocked cues with additive pre-training Method Participants This trial was done by 120 undergraduate university student who were separated into two groups numbering 60 students in each. Of the sample 78 were female and 42 were male. The mean age of students who participated in the study was 21.7 years of age. Procedure All 120 participants sat a computer trial to determine which foods were causing a hypothetical ‘Mr X’s’ allergic reactions. This initially began with a training phase. On a number of trial participant had to predict whether an allergic reaction would occur by click on of two or three options on the screen. Next participants were then told whether an allergic reaction had occurred and how severe the reaction was on a scale ranging from 0 to 10. This was Stage 1 was done by participants 8 times. In stage 2 compounds were also presented 8 times each. Participants could choose from no reaction, or one of two alternative reactions. In stage 2 a reaction occurred could have a severity of either 5 or 10. Next the 20 cues were presented individually and participants were to make two ratings for each cue. First to pair the cue with the allergic reaction that had occurred. The cues A-H were blocked cues, S-Z overshadowed control cues, and K-N no outcome cues. Next they rated their confidence on a scale of 0-100, 0 being not confident at all and 100 very confident. On the same scale they then rated the extent to which that same cue actually caused the reaction on another rating scale. Assumptions on additive logical reasoning were manipulated on two two levels: One group (Additive pre-training) were given training trials before blocking training, which showed the effects of combining to other allergenic food cues. They also received 3 explicit instructions stating that the addition of two causal cues resulted in a more severe outcome. The other Non-additive pre-training group received the same pre-training except that two options that cause and reaction gave the same medium allergic reaction of 5 as each of these options individually. This group also received explicit instructions stating that the addition of two causal cues resulted in an outcome of the same severity. The purpose of this design is to evaluate if scores show a lower causal rating for the blocked cues than the overshadowing cues. Also whether participants were less accurate in recalling the outcome for the blocked cues than for the overshadowed cues. It’s also to understand the effect of pre-training on blocked and control cues. Results The results of the study indicate that some areas of the study produced statistically significant results while when comparing other groups no statically significant data was found. When comparisons across groups were assessed it would appear that cue type had a significant effect on blocking (F (1,118) = 99.596 p Figure 1: mean memory accuracy across groups which show no significant difference of blocking between different groups. Figure 2: mean causal reasoning which show a significant difference between groups in terms of blocking. Discussion Prior to the conduction of these experiments due to the results of previous studies it is hypothesised that both additive and non-additive groups would show reliable blocking in memory for both groups. It is also hypothesised that there will be a weaker causal attribution for blocked cues with additive pre-training. Just like in the research done by lovibond (Lovibond P, 2003) pre-training influenced causal reasoning as blocked cues in the experiment were judged as less causal in the additive group. A significant difference in the strength of the blocking effect between groups. This adds evidence to the idea that deductive reasoning about the additive nature of causal events may contribute to the blocking effect as previously thought by other researchers. However the data does still show that even though significant difference in the strength of the blocking effect between groups Blocking still evident after non-additive pre-training, product of associative learning. Another significant result of the experiment was the data showed a strong blocking effect on memory for correct outcome. The outcomes associated with blocked cues were recalled less well than the outcomes associated with the control cues. Pervious idea about the blocking effect (Mitchell Chris, 2006) had proposed that the effect may be due to a failure to encode the B-outcome during training. The fact that data showed a blocking effect for the correct outcome shows that this may be the case. It also helps to explain why it a blocking effect has still been found even though non-additive pre-training had been part of the experiment. Previous research has shown that there are two separate ways of thinking about the blocking, the process as a result of outcome aditivty or a failure to encode the B-outcome. Our research has shown that blocking might be the result of several quite different mental processes. The fact that blocking was still evident after non-additive pre-training suggests that it may be the product of associative learning. Even when measures are taken to discount outcome aditivity a blocking effect it was still present. This means that it is likely that they both contribute to the blocking effect. This experiment needs to be replicate by other studies to ensure this conclusion is valid. It should be extended to different groups of people not just the limited selection of university students in a more scientific setting then a classroom test. It may be that a group of university student with a mean age of 21.7 may not be the best representative same of the population. It may be the lack of scientific setting effected the degree of seriousness when participant undertook the test. It might also be interesting to evaluate something other than blocking to hypothetical allergic reactions as this may in tale different assumptions about food that may have not been controlled for in this study. However students were randomly allocated to different classes for university at the start of semester so that does help with the reliability of the results. Student were also unaware what condition they were in and what was being tested by the experiment which would help prevent a selection bias . Further studies of blocking would be encouraged to eliminate some of the experimental flaws included in this study. This study does show the room needed for further research into the area of blocking as theories of blocking do not appear to be mutually exclusive but underly an associative process that has not yet been discovered. References Dickinson, Anthony. (1984). Judgement of Act-Outcome Contingency: . Quarterly Journal of Experimental Psychology, 36A 29-50. Evan, L. (2006). The relationship between blocking and inference in causal learning. School of Psychology, University of Sydney. Kamin, L. (1969). Predictability, surprise, attention, and conditioning. Punishment and aversive behavior. Lovibond, P. (2003). Forward and backward blocking of causal judgment is enhanced by additivity. Memory and cognition, 133-142. Mitchell, Chris. (2006). Forward blocking in human learning sometimes reflects the failure to encode a cue–outcome relationship. The Quarterly Journal of Experimental Psychology, 830-844.