The Speed,Accuracy And Trade-off

The accuracy of completing a task has always been the main evaluation index in the large-scale assessment.However,during a variety of task situations,all the indexes indicating the quality of the doers are extremely important,including the correctness of the result as well as the timeliness of the decision-making process.If the response time is ignored,much important information will be lost.Therefore,the precision of results not only should be considered,but the timeliness of the decision-making process also should be taken into consideration in any task situation.The timeliness of the decision-making process should also be regarded as an indispensable index.However,participants have to tradeoff accuracy for speed in practical problems.The different tradeoffs have different effects on accuracy on different items.In the existing psychological measurement research,the ordinary hierarchical model explains the relationship between response time and accuracy through the correlation between ability and speed,but it is not enough.Moreover,the ordinary hierarchical model is based on an unknown trade-off state.So researchers further explain the relationship between response time and accuracy at the item level or the interactional level of the item and the participants from the covariance point of view.However,it cannot reflect the dynamic relationship between response time and accuracy.Some researchers introduced residual time as a covariance into the response model,but over-expanded the effect of time on accuracy.In cognitive experiments,a large number of experimental studies were carried out on the trade-offs between speed and accuracy,and the validity of the SAT model was verified.But cognitive experiments are simple in tasks and do not focus on differences between individuals.This paper mainly focuses on the tradeoff between speed and accuracy.From the evaluation of the tradeoff between participants to the establishment of a new hierarchical model of speed and accuracy tradeoff,four studies are carried out:(1)The first study In order to evaluate the speed-accuracy tradeoff of participants,a series of evaluation indexes is constructed based on response time and accuracy.The item time intensity index is the time pressure that the participant can correctly answer the item.If the item time intensity is greater,the participant needs more time to respond correctly to the item.The speed of the participant index reflects the efficiency of the correct extraction of knowledge nodes.The higher the speed level,the shorter the time to extract the same knowledge node.Finally,the speed and accuracy trade-off index is constructed to evaluate the trade-off states of the participants.Then,the participants can be divided into four different trade-off states by the index.In the real data,the item time intensity index and the fluency of the participant index are calculated.In the meantime,the parameters of the log-normal response time model were estimated by the R package LNIRT.First,it is found that the item time intensity index is a positive skew and is highly correlated with the time intensity parameter of the model.Moreover,the speed index also shows a positive skew and is highly correlated with the parameter of the participants.Finally,the participants can be divided into four categories through the trade-off index,such as good and fast,good but slow,slow and poor,or fast but poor.(2)The second study Grounded in the research of speed and accuracy in cognitive experiment,the manuscript will make the cognitive experiment of the speed-accuracy tradeoff model out of the lab and make it a large-scale assessment of model that can be applied to more complex situations of cognitive tasks by building a psychometric model based on the speed-accuracy tradeoff.The model can combine the advantages of cognitive experimental psychology and psychometric models to achieve a more comprehensive and accurate assessment of the quality of participants.The new psychometric model of parameters can be estimated very stably by simulating data.At the same time,the variables of the model and their relations can be well supported by the real data.Finally,the tradeoff states of the participants can be evaluated from different methods by the speed-accuracy tradeoff model.(3)The third study From the point of the hybrid model,a monotonous SAT hierarchical model reflecting the relationship between response time and accuracy is constructed by integrating SAT model,IRT model and response time model.In the monotonous SAT hierarchical model,the dynamic trade-off between response time and accuracy can be reflected.When time is enough,the monotonous SAT hierarchical model reduces to the ordinary hierarchical model.By simulation,it is found that the model parameters can be estimated well.Meanwhile,the fitting index of the monotonous SAT hierarchical model is better than that of the ordinary hierarchical model and response moderation model in the real data.Moreover,the positive and negative discrimination of residuals time is related to the difficulty of the item.(4)The fourth study The relationship between response time and accuracy is more likely to be non-monotonic.Compared with the non-monotonic relationship,the monotonous relationship may lose some information about time and accuracy.Meanwhile,the difficulty of the item may affect the relationship between response time and accuracy.The monotonous SAT hierarchical model is extended to the non-monotonic SAT hierarchical model based on the speed,ability and compensation.On the simulation data,the parameters of the non-monotonic SAT hierarchical model can be better estimated.Moreover,it is more advantageous than the ordinary hierarchical model for real data fitting.Finally,the ability level non-monotonic SAT hierarchical model is better than the speed level non-monotonic SAT hierarchical model.So the non-monotonic relationship between response time and accuracy are more likely to be affected by the ability level.