Research

13 pages
12 views

‘This Is the Person You Selected’: Eyewitnesses' Blindness for Their Own Facial Recognition Decisions

Please download to get full document.

View again

of 13
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
‘This Is the Person You Selected’: Eyewitnesses' Blindness for Their Own Facial Recognition Decisions
Transcript
  See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/264675493 ‘This Is the Person You Selected’:Eyewitnesses' Blindness for Their Own FacialRecognition Decisions  Article   in  Applied Cognitive Psychology · September 2014 DOI: 10.1002/acp.3062 CITATIONS 6 READS 169 3 authors , including:Anna SaganaMaastricht University 13   PUBLICATIONS   53   CITATIONS   SEE PROFILE Harald MerckelbachMaastricht University 478   PUBLICATIONS   13,422   CITATIONS   SEE PROFILE All content following this page was uploaded by Melanie Sauerland on 20 August 2014. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the srcinal documentand are linked to publications on ResearchGate, letting you access and read them immediately.  ‘ This Is the Person You Selected ’ : Eyewitnesses ’  Blindness for Their Own FacialRecognition Decisions ANNA SAGANA*, MELANIE SAUERLAND and HARALD MERCKELBACH Forensic Psychology Section, Department of Clinical Psychological Science, Maastricht University, Maastricht, The NetherlandsSummary: The aim of the current research was to identify conditions under which choice blindness in facial recognition decisionsoccurs. In  󿬁 ve experiments, participants watched four mock-crime videos and made choices that were either evaluative(Experiment 1) or absolute in nature (Experiments 2a  –  c and 3). When participants were subsequently asked to motivate their choice, they were sometimes presented with choices they had not made. For evaluative decisions, concurrent (27%) and retrospec-tive blindness rates (21%) were relatively low compared with previous studies. For absolute decisions, choice-blindness ratesvaried, depending on when exposure to the manipulated outcome took place (immediate: concurrent 32  –  35%, retrospective 0  –  6%[Experiments 2a  –  c]; 48hours ’  delay: concurrent 68%, retrospective 39% [Experiment 3]). We argue that blindness for facial recognition decisions is more likely for evaluative decisions and for longer intervals between decision and manipulation and also for conditions of increased task complexity, which we interpret in terms of ambiguity. Copyright © 2014 John Wiley & Sons, Ltd. INTRODUCTION In many western countries, eyewitnesses  󿬁 rst identify thesuspect from a lineup at the police station and later con 󿬁 rm their decision during a court proceeding. Thus, at trial, eye-witnesses have to testify in support of their earlier decision.Consider though the situation where the lineup administrator (intentionally or unintentionally) wrote down a different decision than the one made by the witness. Would eyewit-nesses notice a change in their identi 󿬁 cation decision whenthey are later interviewed about it during, for example, a court session? The case of Bernard Maughan (Wolchover,n.d.) is an example of such a case. Speci 󿬁 cally, the eyewit-ness in this case made an identi 󿬁 cation by saying  ‘ I think it  ’ s number six ’  in the presence of his solicitor. However,the administrator wrote down  ‘ I think it  ’ s number seven ’ (who happened to be the suspect) and read back the utteranceto the witness. At that point, neither the eyewitness nor thesolicitor demurred. As a result, the of  󿬁 cial record mentioneda member other than the actual identi 󿬁 ed lineup member.Subsequently, the defendant was charged and released on bail.It was only on the appeal, 2years later, that the solicitor of thedefendant spotted the miscommunication while reviewing theidenti 󿬁 cation tapes. This appears to be an incident of   choiceblindness  in the setting of eyewitness identi 󿬁 cation.Choice blindness refers to the dif  󿬁 culty people have indetecting manipulations of a choice they previously madeand their tendency to confabulate introspective argumentsfor the very choice they did not make (Johansson, Hall,Sikström, & Olsson, 2005). In a   󿬁 rst demonstration of thephenomenon, Johansson et al. (2005) asked participants tochoose which of two female faces they found more attrac-tive. After they had made a decision, participants werepresented with their choice and were asked to explain thereasons behind their decision. Using a magic card trick, threeof the 15 trials were manipulated such that participantsactually ended up with the  non -chosen face. Participantswere blind to 308 (87%) of the 354 manipulated pairs. How-ever, when participants were given a description of the proce-dure and were asked whether they would have noticed suchchoicemanipulations,thevastmajorityofthemwerecon 󿬁 dent that they would have done so. Johansson et al. (2005) termedthis meta-cognitive bias  choice-blindness blindness .Inspired by these  󿬁 ndings Sagana, Sauerland, andMerckelbach (2013) examined choice blindness for eyewit-nesses ’  facial recognition decisions. Speci 󿬁 cally, pedestriansin a European city were engaged in a conversation with twoexperimenters who pretended to be tourists. Shortly thereafter,the pedestrians were asked to identify the two experimentersfrom separate simultaneous photo lineups using a forced-choice recognition decision format. Subsequently, they wereconfronted with their selection and asked to motivate their decision. However, for the second target, the chosen lineupmember was swapped with a previously  non -identi 󿬁 ed mem-ber. Interestingly, 68.3% of the pedestrians failed to immedi-ately report that they noticed the change (i.e.,  concurrently ).Even after the end of the experiment in a post-test question-naire (i.e.,  retrospectively ), 39.8% of the participants wereblind to this identity manipulation. Additionally, Sagana et al. (2013) reported that participants who made an accuratelineup decision were more likely to retrospectively detect themanipulation than participants who made an erroneous recog-nition decision. Hence, superior recognition performance wasassociated with higher detection rates.Although blindness phenomena have been demonstratedrepeatedly, its border conditions have been an under-explored facet. Investigating these conditions is important because it may shed new light on factors that affect choiceblindness in settings other than preference, such as theeyewitness identi 󿬁 cation setting. Drawing from the negativeassociation between recognition performance and blindness,the phenomenon may be conceptualized as a manifestationof weak memory. Indeed, memory deterioration for one ’ sown decisions may foster blindness effects, as demonstratedby Sauerland, Schell, et al. (2013). Nevertheless, other stud-ies showed that participants can be blind to changes in theoutcome of decisions that they themselves made only a fewminutes before the presentation of the manipulated outcome *Correspondence to: Anna Sagana, Forensic Psychology Section, Department of Clinical Psychological Science, Maastricht University, P.O. Box 616, 6200MD Maastricht, The Netherlands.E-mail: anna.sagana@maastrichtuniversity.nl Copyright © 2014 John Wiley & Sons, Ltd.  Applied Cognitive Psychology ,  Appl. Cognit. Psychol.  (2014)Published online in Wiley Online Library (wileyonlinelibrary.com)  DOI : 10.1002/acp.3062  (Hall, Johansson, & Strandberg, 2012; Hall et al., 2013;Johansson et al., 2005). Additionally, research on memoryfor choices suggests that people have an enhanced memoryforapreferredoptioncomparedwitharejectedone,evenwhenthe grounds for this decision are unclear (Dellarosa & Bourne,1984; Mather, Sha  󿬁 r, & Johnson, 2000). Thus, it appears that memory decay cannot fully explain choice blindness.Poor processing of the srcinal targets (Johansson et al.,2005) or the manipulated items (Sauerland, Sagana, &Otgaar, 2013) has also been proven to be an insuf  󿬁 cient determinant of choice blindness. Furthermore, participants ’ compliance (Johansson, Hall, & Sikström, 2008; Johanssonet al., 2005; Sauerland, Sagana, et al., 2013), suggestibility(Merckelbach, Jelicic, & Pieters, 2011; Sauerland, Schell,et al., 2013), or tendency to react in socially desirable ways(Merckelbach et al., 2011; Sauerland, Sagana, et al., 2013)does not seem to modulate the phenomenon. The degree of similarity between the srcinal target and manipulated item has sometimes (Hall, Johansson, Tärning, Sikström, &Deutgen, 2010; Sagana et al., 2013; Sauerland, Schell,et al., 2013), but not always (Johansson et al., 2005), beenfound to act as a moderating factor.From another point of view, choice blindness is related tothe constructive nature of preferences and self-persuasion(Johansson, Hall, & Gardenfors, 2011). Speci 󿬁 cally, Johanssonet al. (2011) argued that people have limited access to thereasons for their actions (also Nisbett & Wilson, 1977). As a consequence, people can come to like what they are told or suggested to like. Endorsing choices suggested by othersmay generate a degree of cognitive dissonance, which peoplewant to overcome (Festinger, 1962; Henkel & Mather,2007). Hence, they deliberate reasons to convince themselvesabout their assumed choice. Once the changed outcome haspublicly been endorsed, people need to convince themselvesthat they, indeed, prefer that outcome. Relatedly, after peoplehave accepted the manipulated choice, they may distort their memory for the selected outcome, through self-deception, inorder to reduce dissonance.Taken together, the literature on choice blindness showsthat people have poor insight into their decisions or   󿬂 exiblepreferences that are easily in 󿬂 uenced by external cues. Either way, the factors that affect the occurrence of the phenome-non remain unclear. Nonetheless, the relevance to legalsettings and the potential devastating consequences of the phenomenon for (innocent) defendants are evident.With these considerations in mind, the experiments inthis paper constitute a   󿬁 rst exploration of the borderlineconditions under which blindness for facial recognitiondecisions materializes.Sagana et al. (2013) demonstrated the relevance of choiceblindness for eyewitness decisions. However, the employed 󿬁 eld study approach allows little control over the speci 󿬁 cconditions that foster blindness phenomena in the eyewitnesssetting. The dif  󿬁 culty of the matter is demonstrated in thereduced blindness rates for recognition decisions comparedwith previous  󿬁 ndings using visual stimuli (Johanssonet al., 2005, 2008). One reason for this discrepancy couldbe the use of absolute versus evaluative decisions. Eyewit-nesses base — or should base — an identi 󿬁 cation decision ontheir memory of the individual rather than on evaluative judgements. This is not to imply that eyewitnesses do not make evaluative decisions in real life. In fact, in an archivalanalysis of persons ’  descriptions, Sporer (as cited by Tuckey& Brewer, 2003) found that about 5% of the descriptorsreferenced perceived personality characteristics. In principlethough, eyewitness identi 󿬁 cations in an experimental settingcan be tested against a ground truth, and their outcome isabsolute (i.e., correct or incorrect). Given the objectivenature of identi 󿬁 cation decisions, one would intuitively ex-pect that manipulations would readily be detected, implyinglittle room for choice blindness. This is very much unlike for evaluative judgments about faces. A person may judgesomeone more or less sympathetic depending on their moodor on the context (Loewenstein & Small, 2007). Accord-ingly, uncertainty about the evaluative decision made at anearlier point may easily occur.If a missing evaluative component is responsible for thelower blindness rates of  Sagana et al. (2013), it is reasonableto anticipate increased levels of blindness when an evalua-tive element is introduced to an eyewitness recognition task.To test this hypothesis, we employed a typical eyewitnessrecognition paradigm with the difference that participants,instead of selecting the perpetrator from a lineup, were askedto indicate which of the actors seen in a mock crime theyfound most sympathetic (Experiment 1). Additionally, toexplore factors that might be critical for the occurrence of choice blindness, we performed a series of three studies(Experiments 2a  – c) where absolute decisions were required.Hence, in Experiments 2a  – c, we employed a proceduresimilar to that in Experiment 1, with the difference that participants made a facial recognition decision rather thansympathy evaluations. We anticipated lower blindness ratesfor the absolute compared with evaluative decisions. Exper-iments 2a  – c also differed in the number of manipulations,mask duration (interval between 50 and 500milliseconds),and sample characteristics in order to further investigateconditions that facilitate or inhibit blindness. Finally, in a  󿬁 fth study (Experiment 3), we examined whether increasedretention intervals between the recognition decision and thepresentation of the manipulated outcome would affect blind-ness rates. We considered this important as it is unlikely that the identi 󿬁 cation procedure and the confrontation with one ’ sdecision at trial take place on the same day (Shermer, Rose,& Hoffman, 2011). Thus, Experiment 3 employed a proce-dure similar to that in Experiment 2a with the exception that the manipulated outcome was presented 48hours after therecognition task, which is a delay that adds to the ecologicalrelevance of the study. EXPERIMENT 1Method Participants Thirty-four participants (19 men and 15 women) of variousnationalities took part in the study (  M  age =25.2years,  SD age =5.2, range: 19 – 45). The majority (61%) were Maastricht University students studying different majors (law: 14.7%,psychology: 11.8%, and mental health: 11.8%), whereas38.2% were employees (private sector: 23.5%, public sector:A. Sagana   et al. Copyright © 2014 John Wiley & Sons, Ltd.  Appl. Cognit. Psychol.  (2014)  2.9%, and other: 11.8%). All reported studies were approvedby the standing ethical board of the Faculty of Psychologyand Neuroscience of Maastricht University. Materials Stimulus  󿬁 lms Four stimulus video fragments showing minor offenses weredisplayed on a computer screen. The mean duration was187seconds (duration range: 160 – 214seconds). Two of thefragments showed the theft of a wallet at a university cafete-ria (Cafeteria 1 and Cafeteria 2). The third fragment depictedthe theft of a wallet in a bar (Bar), and the fourth one wasabout an assault attempt at a bus stop (Bus). In each videofragment, there were four actors: a perpetrator, a victim,and two bystanders. Actors were different for each  󿬁 lm.The gender of perpetrators and victims was counterbalancedacross videos. In two videos (Cafeteria 1 and Bus), theperpetrator was a man and the victim a woman, whereas thereverse was true for the other two videos (Cafeteria 2 and Bar).Each target was in sight for a minimum of 76seconds, withclose-ups of 2 – 9seconds. All targets were shown from frontaland side views. Post-test questionnaire To examine whether participants had noticed our manipula-tions but refrained from mentioning it, we administered a post-test questionnaire that was adjusted from Johanssonet al. (2008). Participants were  󿬁 rst asked whether they hadnoticed anything strange during the experiment, and whenthey responded af  󿬁 rmatively, they were invited to providedetails. If they reported they noticed the manipulation, theywere coded as retrospective detectors. Next, participantswere given a description of an imaginary experiment inwhich some identi 󿬁 cations were manipulated in such a waythat the participant would end up with a photo that she or he did not choose. To test choice-blindness blindness(i.e., the meta-cognitive bias; Johansson et al., 2005), partic-ipants  󿬁 rst had to indicate if they thought that they wouldnoticesuch achange.Second,they wereaskedifthey believedthat we had carried out such a manipulation in the current experiment. If participants responded af  󿬁 rmatively to thisquestion, they were again counted as retrospective detectors. Design In Experiment 1, we manipulated the preferred target in oneof the four video fragments. Speci 󿬁 cally, the decisions of theBus or Cafeteria 2 fragments were manipulated. Detectionrates were measured both  concurrently  and  retrospectively .Concurrent detection refers to participants who immediatelynoticed the manipulation after it had taken place. Retrospec-tive detection, next to concurrent detection,  additionally includes those participants who reported in the post-test questionnaire that they had noticed the change. In accor-dance with Johansson et al. (2008), any positive answer tothe questions about noticing anything strange or noticing a manipulation taking place during the experiment wascounted as retrospective detection. Procedure All parts of the experiment were presented on a 20-in. com-puter screen at a resolution of 1024×768pixels usingDMDX Display Software (Forster & Forster, 2003). Partici-pants were naïve to the actual purpose of the study and weretested individually.After signing the informed consent form, participantswere told that they would watch four video fragments andwould act as eyewitnesses. Then, the  󿬁 rst video waspresented. After the end of the fragment, participants wereshown a slide with photos of the actors they had seen duringthe video. Their task was to select which of these actors theyfound most sympathetic. Although the procedure is verysimilar to that in an identi 󿬁 cation experiment, participants ’ task was evaluative in nature.For the three video fragments (Cafeteria 1, Cafeteria 2,and Bar) in which three actors were of one gender and onlyone actor of the other gender, we presented only the threeactors of the same gender (in one row of three pictures). Thiswas carried out to prevent an increased detection rate simplydue to noticing the gender difference. For the fourth video, inwhich two male and two female actors were shown, all four actors were presented in two rows of two pictures.After participants had made their choice, a 50-millisecondrandom pattern (i.e., mask) was displayed, followed by thepresentation of the chosen actor. Now participants wereasked to justify their decision. The same procedure wasfollowed for all four videos that were presented either inthe order Cafeteria 1, Bar, Cafeteria 2, and Bus or in theorder Cafeteria 1, Bar, Bus, and Cafeteria 2. However, thedecision for the video fragment presented last (either Busor Cafeteria 2) was manipulated, and participants wereconfronted with a predetermined photo they had actually  not  selected. If participants ’  justi 󿬁 cations included a comment in-dicating that the displayed photo did not correspond to their choice or that the program had made a   ‘ mistake ’ , they wereclassi 󿬁 ed as concurrent detectors. Detectors frequently alsoverbally informed the experimenter about the change. No sig-ni 󿬁 cant differences in detection emerged between the two or-ders for concurrent,  x  2 (1,  N  =34)=1.89,  p =.25,  phi =0.24,or retrospective detection,  x  2 (1,  N  =34)=2.10,  p =.21,  phi =0.25. Therefore, we will not discuss this factor further.At the end of the experiment, participants were given thepost-test questionnaire and were fully debriefed. Results and discussion The concurrent detection rate was 73.5% ( n =25), and retro-spective detection was 79.4% ( n =27). On the whole, 20.6%( n =7) of the participants were completely choice blind. Of these choice-blind participants, 85.7% ( n =6) said in thepost-test questionnaire that they would be able to detect themanipulation. Thus, this group evidently exhibited choice-blindness blindness.Theobservedconcurrentandretrospectivechoice-blindnessratesofthepresentstudyarerelativelylowcomparedwithpre-vious choice-blindness studies for visual stimuli (Johanssonet al., 2005, 2008). This may be due to the use of a videofragment with a speci 󿬁 c plot. Speci 󿬁 cally, it seemed that participants based their preference on the role of the target Witnesses ’  blind facial recognition decisions Copyright © 2014 John Wiley & Sons, Ltd.  Appl. Cognit. Psychol.  (2014)  and their empathic feelings toward the victims and thebystanders. This is illustrated by the low preference ratefor the perpetrators (3 of the 34 participants) compared withthe victims and bystanders (91.2%). Thus, because their decision was often based on a clear category (e.g., victims),participants may have found it relatively easy to detect themanipulated outcome.Most importantly and in contrast to our hypothesis, thechoice-blindness rates observed in Experiment 1 are lower compared with those reported by Sagana et al. (2013;31.7 – 60.2%). These  󿬁 ndings do  not   speak to the idea that blindness rates depend on the decision type required (i.e., ab-solute vs. evaluative). The diversion in blindness rates acrossthe two studies could be due to differences in methodology,as the present is a laboratory study, whereas Sagana et al.(2013) performed a   󿬁 eld study. That is, the variousdistracters that are inherent to real-life interactions may haveincreased the complexity of the task, resulting in raisedblindness rates (Sagana et al., 2013). To test this notion,we examined whether choice blindness would be lessprominent if participants performed a procedure similar toExperiment 1, however with participants performingabsolute facial recognition decision rather than evaluativedecisions. To this end, we performed a series of three studies(Experiments 2a  – c). In addition to varying decision type(compared with Experiment 1), Experiments 2a  – c differedin the number of manipulations, mask duration, and samplecharacteristics in order to test other conditions that couldcon 󿬁 ne blindness for facial recognition decisions. EXPERIMENTS 2A – CMethod In Experiments 2a  – c, participants made forced-choice recog-nition decisions from target-present lineups. In Session 1,participants watched four video fragments and made facialrecognition decision for each of the four targets depicted inthe  󿬁 lms (i.e., 16 targets in total). After each decision, a mask was presented, and subsequently, the photo of the selectedlineup member was displayed on the screen. However, somedecisions (either two [Experiment 2a] or four [Experiments2b and c]) were manipulated and replaced with a non-selected lineup member. Twenty-four hours later, partici-pants returned to the lab for a second session and were askedto identify all perpetrators and victims again (i.e., eight targets in total). Session 2 was introduced to investigatewhether the manipulations had an impact on participants ’ future decisions.For Experiments 2a  – c, the same stimulus  󿬁 lms, lineups,testing materials, and procedure were used. In Experiments2aandb,wetestedwhetherblindnessforfacialrecognitionde-cisions was affected by the number of manipulations and bymask duration. Speci 󿬁 cally, we decreased the number of ma-nipulations from four in Experiment 2a to two in Experiment 2b and increased mask duration from 50 to 500milliseconds.This was to test whether a large number of manipulations aswell as the short mask duration would alarm participants andhence increase detection.In Experiment 2c, we investigated whether our resultswere affected by our heavy reliance on bachelor psychologystudents as participants. Students may just have waited for some kind of manipulation to occur while they were tested,or they might have read or heard about the choice-blindnesseffect during the course of their studies. Relatedly, the useof undergraduate student samples has been vigorouslycriticized among others as malleable to social in 󿬂 uences(e.g., Sears, 1986). To rule out this possibility, Experiment 2c relied on participants who were currently not active in or had never been involved in academia. The procedure of Experiment 2c was analogous to that of Experiment 2b,with the exception that there was only one session. Thisseemed appropriate because we were mostly interested inthe blindnessrates inSession 1. Additionally, we asked partic-ipants in the post-test questionnaire how many manipulationsthey had detected in Experiment 2c. Participants In Experiments 2a and b, participants were 18 and 19 (26women) students of Maastricht University, respectively(  M  age =20.9years,  SD age =1.6, range: 19 – 25). The majoritywas psychology students (73.0%), and one was a mentalhealth student (2.7%), whereas the majors of the remainingstudents (24.3%) were not speci 󿬁 ed. In Experiment 2c, 20(11 women) participants (  M  age =41.5years,  SD age =16.2,age range: 19 – 64) had varying professional backgrounds(private or public sector: 35%, freelance: 20%, household:20%, and temporary employment: 25%). Participation wasvoluntary. Student participants received course credit inreturn, whereas for non-academics, no monetary or other incentive was granted. All participants were naïve to thepurpose of the study and were tested individually. Materials  Lineups A total of 16 simultaneous, target-present photo lineups werecreated. To that end, head-and-shoulder photos were selectedto match the description of the respective targets. Eachlineup was presented on a 2×3 array format. Thus, eachlineup included one target (i.e., perpetrator, victim, or bystander) and  󿬁 ve distracters. The size of the photos, aspresented on the computer screen, was 9.0cm×10.0cm.Only target-present lineups were administered, and partici-pants were not allowed to make rejections. We opted for thisprocedure because we wanted to test for the existence of choice blindness for lineup selections (rather than rejections).Note that other phenomena such as the feedback effect (Wells& Brad 󿬁 eld, 1998, 1999) or the own-race bias effect (for a complete review, see Meissner & Brigham, 2001) weresrcinally studied using a similar approach.  Design In Experiments 2a and b, we manipulated either four or two of the 16 lineups. Table 1 summarizes the manipulations per experiment. Detection rates were measured both concurrentlyand retrospectively.A. Sagana   et al. Copyright © 2014 John Wiley & Sons, Ltd.  Appl. Cognit. Psychol.  (2014)
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x