Eyewitness testimony
Tutorial
Problem 1B:
1)
Arie mentions new details but also forgets previous details about looks of suspicious men
Potentially he spoke to others about it, for instance to his brother, so you might fill in gaps or
recollect other things.
Perhaps he’s thinking that he shouldn’t mention the things he already mentioned.
2)
How accurate (consistent!!) will statements be directly after and month after an incident
80-90%
70%
3)
Should IND use inconsistencies to estimate trustworthiness?
Memory is flexible, also depends on cues you give. If something is traumatic it can come in parts,
which doesn’t imply not trustworthiness.
Language barrier should be taken into account.
Learning objectives:
1) Definitions of
- What is accuracy?
- What is consistency?
- What is completeness?
- How are these three terms related??
2) What is the influence of time on
a) accuracy
b) consistency
c) completeness?
3) Does inconsistency in testimony imply inaccuracy?
Literature
PROBLEM 1A
Ackerman – Control over grain size in memory reporting- with and without satisficing knowledge
Different levels of granularity – precision or coarseness (grofheid)
This article deals with process of answering questions from memory.
1. Grain size
= level of details of information that can be assessed
Depend on the operation of memory encoding, retrieval and reconstruction processes
Mainly: passage of time
2. Social-communication approach:
People’s responses to questions take into account pragmatic considerations and tacit (stilzwijgende)
assumptions
,Responses to questions take into consideration:
- Background and existing knowledge of the questioner
- Purpose in asking the question
- Personal goals
- Self-expectations
3. Metacognitive approach
Shares the emphasis of the social-communicative approach on strategic behavior guided by personal
and social goals, while at the same time attempting to specify the monitoring and control processes
that underlie strategic memory performance and integrate these processes into memory theory.
Memory performance (quantity and accuracy) depends on interaction between
(1) Contents of memory
(2) Metacognitive processes that guide memory search and retrieval, convert retrieved
information into concrete candidate answers, determine whether the answers are reported
and in what manner
Research question present study:
How people regulate the grain size (precision or coarseness) of answers to memory questions.
Control of grain size in memory reporting is guided by two main objectives
(1) Correctness
(2) Informativeness
When one is uncertain about correctness of one’s memory, the two tend to conflict:
To increase chance that answer is correct, coarseness of answer should be increased, by doing so
however; the answer becomes less informative and will generally be less appreciated by the recipient
of the information.
Previous studies -> Grain-control model:
Emphasizes the aim of correctness over the aim of Informativeness (people rather provide
coarseness answers which are more likely to be right and provide less information), incorporating the
goal of correctness by way of an explicit confidence criterion, whereas the goal of Informativeness is
handled in a more indirect matter.
This study:
Dual-criterion model is needed to explain the choice of grain size when knowledge is relatively poor.
Satisficing model – Goldsmith:
Respondent strives to provide as much information as possible, as long as its subjective probability of
being correct satisfies some reasonable minimum level.
The level of this minimum-confidence criterion is assumed to depend on the relative incentives for
correctness and Informativeness in answering a particular question in a particular social situation.
Correctness emphasized -> higher criterion
Informativeness emphasized -> lower criterion
Relative expected-utility maximizing model – Goldsmith:
Respondent calculate the subjective expected utility of candidate answers at various grain sizes,
(taking into account both the subjective value of a correct or incorrect answer, and the subjective
probability that the answer is correct or incorrect) compare these values and ultimately choose the
answer shoe subjective expected utility is maximal.
Questions were presented in two phases
, (1) Participants answered each item using two different bounded intervals specified by the
experimenter
Participants also rated their confidence in the answer at each grain size by assessing its probability of
being correct (exp. 2 & 3)
(2) Participants went over their answers and for each item indicated which of the two grain sizes
they would prefer to provide assuming that
a. they were an expert witness testifying before a government committee (exp. 1 & 2)
b. they earned monetary payoffs, which were larger for correct fine-grained answers than
for correct coarse-grained answers (exp. 3)
Results: support simple satisficing model
The participants tended to provide the coarse-grained answer when the fine-grained answer had a
low subjective probability of being correct.
The grain control decision was based primarily on confidence in the fine-grained answer, as predicted
by the satisficing model, and not on the relative disparity between confidence in the precise- and
coarse-grained answers, as predicted by the relative expected-utility maximizing model.
Limitation of simple satisficing model,
If someone absolutely does not know, the model states that someone will produce a ridiculously
coarse answer to be reasonably sure that the answer is correct.
There is research that states that people do not
- Yaniv & Foster : only 47% of the participants’ answers were correct, and in fact, the provided
intervals would need to be widened across the board by a factor of 17 to achieve the
specified hit rate of 95%
Reasons could be
(1) Overconfidence – people may have believed that their answers were coarse enough to
achieve a very high accuracy rate
(2) Reluctance to provide extremely coarse answers that violate social norms of communication,
specifically, the expectation that one’s answers should be reasonable informative
Thus, the influence of pragmatic communication norms that constrain the coarseness of socially
acceptable answers “may diverge from and even supersede the demands of calibration accuracy
Dual-criterion model:
The informativeness criterion reflects the minimum level of precision, or maximum level of
coarseness, that is perceived as constituting a reasonably informative answer to the question.
The minimum-informativeness criterion should depend on such factors as the perceived expectations
and needs of the questioner in asking a particular question, the respondent’s own self-expectations
concerning the question and social norms for particular types of questions in particular contexts.
The Informativeness criterion is set after each question has been presented but before the answering
process begins.
Two theoretical knowledge states
(1) Satisficing knowledge SK
= when level of knowledge is sufficient to allow one to provide an answer that simultaneously
satisfies both confidence and Informativeness criteria.
If SK then there is in fact no difference between dual-criterion model and original satisficing model.
(2) Unsatisficing knowledge UK
During the question answering process, then, respondents in a UK state will eventually reach a
deadlock situation: If they begin with a precise answer (e.g., a best guess) and attempt to coarsen the
answer so that it passes the confidence criterion, the process will be halted by failure to meet the
informativeness criterion before reaching the confidence criterion. Same if visa versa.
Two different ways to resolve this
, 1. To circumvent the criterion conflict entirely by responding to do not know instead of
providing a substantive answer
2. If ‘don’t know’ is denied, one of the grain selection criteria (primarily confidence) will have to
be violated
Hypothesis: Respondents in a UK conflict state will often violate the confidence criterion in deference
to the informativeness criterion.
In the experiments participants were allowed full control over the grain size of their answers. Level of
knowledge was manipulated
- By comparing performance on very hard versus moderately hard sets of items
- By pre-exposing the answers to half of the very hard items
To see if
(1) Participants will use unlimited control over grain size to achieve same level of correctness for
very hard and moderately hard items (statistic model)
(2) Participants will give low-confidence answers to achieve minimum level of informativeness
(dual-criterion model)
Results
- Accuracy of answers to the low knowledge questions was lower than the accuracy of
moderated knowledge answers.
- Overconfidence bias -> participants were overconfident in the correctness of their answers,
particularly for the low knowledge questions
- Confidence criterion is sometimes violated to maintain a reasonable level of informativeness,
if primarily knowledge is low -> support for dual-criterion model
- Participants use don’t know response to resolve criterion conflicts primarily when their level
of knowledge was low
- Pre-exposing to answers did not change answer patterns
Conclusion:
Even if participants know very little, they choose not the most coarse answer
Informativeness goes before confidence
Evans & Fisher – Eyewitness memory: balancing the accuracy, precision and quantity of
information through metacognitive monitoring and control
Input bound measure of memory = measuring performance based on the quantity of information
recalled = proportion of presented items that are later recalled
Using this, performance declined as retention interval increased -> ‘forgetting curve’
This makes the assumption that 1) the long delay since the crime caused the witness’s memory to
decline and 2) because of this decline the accuracy of the witness’s memory is questionable.
Output bound measure of memory = measure the accuracy of recollections = proportion of
responses provided that are correct
Accuracy rate = output bound measure, the proportion of all witness statements that are accurate.
Retention interval = the period between a participant’s exposure to information and being tested for
retention of that information
What explains why retention interval might only have minimal or no effects on accuracy of witness’s
report?
1. Report option = the opportunity to withhold a response
