Simultaneous VS Sequential

Simultaneous VS Sequential

Terminology

Hit (True positive) = when a guilty person is identified as guilty

Miss (False negative) = If a guilty person is seen as innocent

False alarm (False Positive) = If a innocent person is identified as guilty

Correct rejection (True Negative) = if line up with innocent people is rejected

Diagnosticity Ratio

If a suspect is identified how likely is it that the suspect is guilty

Diagnosticity = Hit rate/False Alarm

We want more hits and fewer false alarms

Procedures with larger diagnostcity are considers the best procedures

Diagnosticity - Sim vs Seq

Meta analysis of 72 test

Sim

Hits = 52

False alarms = 9

Diagnosticity = 5.8

Seq

Hits= 44

False alarms =5.6

Diagnosticity = 7.8

According to this sequential is better, studies that report diagnosticity ratio often reveal a sequential advantage

Signal detection theory

Discriminability: ability to distinguish between a signal and noise

Line up discrimination

Signal = culprit

Noise = innocent line up members

Need person to identify the signal and not be distracted by the noise

Response bias: inclination to accept that the signal is present

Lineup response bias: willingness to make an ID

Liberal = more willing to choose

Conservative = less willing to choose

Sequential is more conservative response bias

Simultaneous is more liberal response bias

Signal Detection Theory - Criticism

Doesn’t exactly measure discriminability

Dignosticity ratio is not a pure measure of discriminability

Diagnosticity increases as responding becomes more conservative

Despite being the same difference between hit and false alarms dignositicity is increasing when people are less willing to choose

Receiver Operating Characteristic (ROC) Curve

Signal detection theory suggests ROC curves as a type of analysis other than dignositicity

Plotted at how many hit rates you have at each level of confidence

E.g.

O% confidence = HR = 2 FA = 5

Calculate using AUC – area under the curve

Larger area under the curve means it’s the better procedure

ROC study

All experiments by Mickes, Flowe and Wixted 2012 found that simultaneous was better

ROC findings

Studies using ROC curves tend to find that simultaneous advantage

Comparisons between faces enable witnesses to find features that distinguish the culprit from the fillers

Criticisms of ROC

1 ROC does not measure underlying (psychological) discrimination

Discriminability: Ability to distinguish a signal from noise

Signal: Culprit

Noise: Innocent Lineup Members

For lineups with innocent suspects, ROC analysis treats a filler identification as if it were a lineup rejection

ID Filler = Incorrect Decision

Reject Lineup = Correct Decision

Eyewitness identification have six possible outcomes

But ROC curves are designed for decisions that have four possible outcomes

2 ROC does not control for response bias

Response bias: Willingness to accept that a signal is present

ID = Believe the signal is present

Rejection = Believe the signal is absent

ROC analysis combines filler identification and lineup rejections

Conclusion about research

Superiority of simultaneous/sequential procedures may depend on how performance is measured

§  Diagnosticity: Advantage for sequential

§  ROC Curve: Advantage for simultaneous

Consensus is that witnesses are more conservative for sequential than simultaneous

Future of Lineup ID

Immersive stimuli

VR and 360 may produce more naturalistic results

One problem is that VR environment is hard to create and requires a lot of funding

Hybrid Lineup

Combines both types of lineup

All options are presented simultaneously but movement is presented sequentially

Some initial evidence that is more diagnostic than other procedures but results still unpublished

Trying to combine the best of both ideas

Interactive Lineup

Pose reinstatement

More likely to recognise someone if they are in the same pose they were when you saw them

o   The procedure allows witnesses to the lineup to view the faces from multiple angles.  The procedure also allows the user to control the images, and as such is interactive.

There is evidence that the more unique features you see, the more accurate your ID will be.

Interactive lineups improve ability to discriminated between innocent and guilty suspects

For possible false ID rate, interactive lineups increased the correct ID rate by 18%

Culture

South Korea vs UK

Sequential procedures had higher diagnosticity among UK participants, but not among SK participants.

Composites - Police Sketches

Police sketch artists were used in as many as 10% of law enforcement agencies in the mid-1980s (MacDonald, 1984)

Composites - Mechanical systems

The Identi-Kit and the Photofit Kit

use transparencies of facial features that are superimposed over each other to create facial likenesses

§  Hundreds of hairstyles and eyes, dozens of mouths, noses, and chins are available for the person to select when building the face

Composites - Computerised systems

One of the computerized systems in use in U.S. law enforcement agencies today is the FACES.

§  The person must select individual facial features and combine them for image of a face.

§  Called Feature based systems – as you select individual facial features

Composites - How Good

o   Initial results were disappointing (e.g., see Davies, Ellis, & Shepherd, 1978), and modern computerized versions have not done much better (Davies & Valentine, 2006).

Participants correctly named only 22 of 800 famous faces based on composites of those faces that had been created by other individuals.

o   It has been argued that there is a mismatch between the task demands that are somewhat inherent in all composite systems and the way that faces are usually perceived and remembered.

§  Faces are generally processed, stored, and retrieved at a holistic level rather than at the level of individual facial features (Tanaka & Farah, 2003).

Holistic-based systems

EvoFIT

§  A random set of faces is generated, and the witness selects the face or faces that are most similar to memory for the target face.

·       Based on some descriptive factors

§  A new set of faces is then generated, based on mutations of the previous selections, and the witness again makes a choice.

§  This process is repeated until the witness cannot choose because all the faces resemble the target face equally well.

o   Results of these

§  Better results with holistic approaches for matching and therefore is a better procedure

Morphing facial composites

o   Some research has shown that there can be benefits to morphing multiple composite faces of the same individual created by different eyewitnesses

§  (Bruce, Ness, Hancock, Newman, & Rarity, 2002; Hasel & Wells, 2007).

Limitation - Composites

Perceptions of facial composite similarity are susceptible to bias

§  Detectives and jurors may use the perceived similarity of a suspect to a composite as evidence to determine the likelihood of a suspect’s guilt.

§  However, these similarity judgments may be biased by evaluators’ preexisting beliefs of guilt (Charman et al., 2009).

Facial composites may distort memory

§  Participants either constructed a facial composite or completed a distractor task.

·       Those who created a facial composite had lower identification accuracy in a line-up identification task.

·       “Creating a composite or simply viewing someone else’s composite hinders identification accuracy and one’s memory for the target face”