Face Recognition
Photofit/identikit- one technique for getting eyewitnesses to recreate an image of the face, however has been proven not very effective
Brain is specially developed to recognize faces and tell people apart, much harder to do if they are upside down
Martha Farah of the University of Pennsylvania proposed we recognize objects via 2 mechanisms housed in different areas of brain - first mechanism is via the strategy of decomposing the object into its component parts or features. Second object recognition involves more holistic approach
Photofit systems are focused on getting people to analyze faces feature by feature, usually are very poor because research indicates brain does not recognize faces feature by feature but by the whole view
Evofit- New technique where eyewitness chooses between 72 random faces 6 that look most like person, then generates 72 new faces based on the 6 they selected to pick 6 more, this was found to be accurate around 25% of the time while using feature based photofit was only accurate 5% of the time
Evofit purposely blurs external features such as neck and ears to help focus on internal such as eyes and nose, you are able to adjust features but only from a wholistic approach
Single-blind procedure - where only the eyewitness doesn't know in advance who the suspect is, this can potentially be a problem because the administrator may accidentally provide cues to the eyewitness as to which person is the suspect
Double-blind procedure - where both the admin and eyewitness do not know who the suspect is
This procedure leads to better outcomes
Biasing behaviors - can be shown through administrator, especially when single-blind procedure is used even if suspect is innocent, witness believes 90% of time the person committing crime is in line-up, despite if they actually are, began reminding them they may or may not be in line-up
Response Bias- Bias instructions can be both good and bad
Never know in real life, should always use unbiased instructions so there is less chance of sending an innocent person to prison
Should pictures be laid out for witness side by side, this approach may encourage relative rather than absolute judgements
sequential line up - Roderick Lindsay of Queens University, where each person is shown one by one the photo, then have no idea how many there will be, asking did you see this person at the crime?
Flaw in this one is lack of comparison but still results in fewer false identifications
Can be helpful to first show "blank line-up" where no suspect is present to see if there information may not be accurate
Foils- members of the line-up who are not the suspect, should all be wearing similar clothing and overall look similar, yet if they look too similar they will not be able to tell which is the suspect
Indicator of eyewitness accuracy - Confidence! Some argue this is more important factor
Correlation Coefficient- a statistic that measures the strength of a relationship between two variables
Photofit/identikit- one technique for getting eyewitnesses to recreate an image of the face, however has been proven not very effective
Brain is specially developed to recognize faces and tell people apart, much harder to do if they are upside down
Martha Farah of the University of Pennsylvania proposed we recognize objects via 2 mechanisms housed in different areas of brain - first mechanism is via the strategy of decomposing the object into its component parts or features. Second object recognition involves more holistic approach
Photofit systems are focused on getting people to analyze faces feature by feature, usually are very poor because research indicates brain does not recognize faces feature by feature but by the whole view
Evofit- New technique where eyewitness chooses between 72 random faces 6 that look most like person, then generates 72 new faces based on the 6 they selected to pick 6 more, this was found to be accurate around 25% of the time while using feature based photofit was only accurate 5% of the time
Evofit purposely blurs external features such as neck and ears to help focus on internal such as eyes and nose, you are able to adjust features but only from a wholistic approach
Single-blind procedure - where only the eyewitness doesn't know in advance who the suspect is, this can potentially be a problem because the administrator may accidentally provide cues to the eyewitness as to which person is the suspect
Double-blind procedure - where both the admin and eyewitness do not know who the suspect is
This procedure leads to better outcomes
Biasing behaviors - can be shown through administrator, especially when single-blind procedure is used even if suspect is innocent, witness believes 90% of time the person committing crime is in line-up, despite if they actually are, began reminding them they may or may not be in line-up
Response Bias- Bias instructions can be both good and bad
Never know in real life, should always use unbiased instructions so there is less chance of sending an innocent person to prison
Should pictures be laid out for witness side by side, this approach may encourage relative rather than absolute judgements
sequential line up - Roderick Lindsay of Queens University, where each person is shown one by one the photo, then have no idea how many there will be, asking did you see this person at the crime?
Flaw in this one is lack of comparison but still results in fewer false identifications
Can be helpful to first show "blank line-up" where no suspect is present to see if there information may not be accurate
Foils- members of the line-up who are not the suspect, should all be wearing similar clothing and overall look similar, yet if they look too similar they will not be able to tell which is the suspect
Indicator of eyewitness accuracy - Confidence! Some argue this is more important factor
Correlation Coefficient- a statistic that measures the strength of a relationship between two variables
Neil Brewer of Flinders University and Gary Wells describe calibration techniques that allow us to calculate how over or under confident an eyewitness is for a range of different accuracy levels, in controlled situations, make chart looking at confidence and under, used to determine likelihood if identification is accurate
We need to work on either the selected template or the set up. It is hard to read and the format is off.
ReplyDelete