I was a cognitive psychologist interested in induction, memory, and perception. My current research is on consciousness and unconscious processing, and prototype formation using binary sequences. I am also a survivor of a concentration camp. After retirement my interest in AII, protoype forrmation, and consciousness increased. I am also editor of an educational technology journal, Interactive Learning Environments. Supervisors: W. R. (Tex) Garner, Endel Tulving, and Bruce Henning Address: Gaithersburg, Maryland, United States
These memory experiments follow Psotka et al. (2022 b) with distorted binary patterns and examine... more These memory experiments follow Psotka et al. (2022 b) with distorted binary patterns and examine how chunking improves memory for binary patterns, and how many repetitions are needed for the first correct prediction. A simple paradigm is used, where less than 1 element of each base binary pattern of length 16 is randomly complemented, on average, and presented in a list sequences, presented discretely and sequentially. Each sequence is presented one element at a time and the next element is guessed. After each presentation of all 16 elements (onre presentation) the Experimenter/Subject guessed how many distortions were in the pattern Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication. On each trial, the S/E chunked the sequences into subsequences of length 4. This conscious rule arose after many trials and was tested in a separate experiment. Chunking dramatically increased the learning rate but learning was still a function of the redundancies in the sequence.
These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and a... more These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and add unconscious creation of prototypes as measured by the Syntely algorithm. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. Here we examine the number of changes or distortions in binary patterns of length 16. A simple paradigm is used, where about 1.5 to 4 elements of a binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What is the learning rate as measured by the Syntely algorithm (correlations between predicted elements versus test elements of the base pattern for sequences where the average number of changes vary)? This experiment then focuses on the easier patterns, recognizable consciously within about 200 trials to look at the Syntely algorithm's rate of learning. Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and a... more These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and add conscious verification by the experimenter and learning rate over 100 trials. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. Here we examine the number of changes or distortions in binary patterns of length 16. A simple paradigm is used, where about three to 6 elements of a binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What is the learning rate (correlations between predicted number of changes and actual changes)? Previous experiments show that the base pattern is selected way above chance with 4 to 16 elements. This experiment focuses on the easier patterns, recognizable consciously within about 200 trials to look at the rate of learning. Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
In an annual review of memory Kahana (2020) argued that ", our core theoretical understanding of ... more In an annual review of memory Kahana (2020) argued that ", our core theoretical understanding of human memory is still firmly grounded in the list recall method pioneered by Ebbinghaus and Müller in the late nineteenth century". Perhaps it is time to change the focus to inductive memory. Ebbinghaus also conducted other experiments on association memory in which he proposed and found evidence that association memory is essentially unlimited, although weaker with distance. This may now be seen as the basis for convolution and distributed memory, which in turn is the basis for factor analysis, regression, and modern artificial intelligence computations. Our experiments are the beginning of a series of monographs exploring binary sequences to confirm and extend Ebbinghaus' findings to inductive memory. This paper, first prepared in 1978, argued that another paradigm must be added to the standard view of memory: Inductive memory (Psotka, 1975) as an extension of association and convolution. Binary sequences of length eight were
These prototype abstraction experiments follow Posner and Keele (1970) and Garner (1974) with bin... more These prototype abstraction experiments follow Posner and Keele (1970) and Garner (1974) with binary patterns and add conscious verification by the experimenter. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. A simple paradigm is used, where about three to 7 elements of each binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What are the strengths and limitations of our ability to discern the base pattern? The task is to select the most representative pattern in the set when it occurs within the set. Previous experiments show that the base pattern is selected way above chance with 4 to 10 elements. This experiment extends the number of elements to 16 and explores the main factor, the mean number of perturbations, that measures the similarity to the base pattern (which is not trained explicitly). Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
Planton et al. (2021) conducted a very thorough set of experiments about memory for binary sequen... more Planton et al. (2021) conducted a very thorough set of experiments about memory for binary sequences, comparing a broad variety of different models. Since memory is fundamental to prediction, I decided to examine their first experiment with 16 elements in detail. Besides, they were one of the very few to cite my work (even though the citation was a little wrong), Psotka (1975). So, I decided to replicate their Experiment 1 to see how much of their memory results could be explained by the prediction components. The Syntely measure I use is a kind of association, autocorrelation, or convolution, with the most recent items used as a recall cue. It also uses a kind of chunking of runs at the terminal part of the sequence when used as the recall cue (cf. Psotka, 1975). Since their paper focuses on the power of recursion over association, I thought this would be a useful comparison. Fortunately, they use a kind of prediction paradigm with reaction time (and error rates) to predict tones after the sequences have been learned, and this fits the Syntely paradigm perfectly. They acknowledge that Experiment 1 is the most informative experiment in this paper because: "… we predicted that the effect of LoT complexity in the proposed language of thought would increase as the sequence gets longer." Cf. p. 43
The After Action Review (AAR) process provides a powerful methodology that in the context of trai... more The After Action Review (AAR) process provides a powerful methodology that in the context of training maximizes the benefits of exercises by enabling a unit to learn from experience by systematically reflecting on their strengths and weaknesses. We have developed a tool that supports the AAR process, essentially extending an Observer Controller's (O/C) reach automatically. This tool was developed with two training contexts in mind: live STX lane convoy training at the National Training Center (NTC) and simulated convoy training using DARWARS Ambush! at the Mission Support Training Facility at Fort Lewis. At NTC, live radio communication is captured during training, while with Ambush! communication using voice over IP (VOIP) is recorded. The tool automatically converts recorded speech to text and then analyzes the text, using advanced statistical machine learning technologies, to determine a unit's performance and identify critical incidents, leading indicators, and other t...
ILE has new functionality that we, as the Co-Editors of this journal, are asking you to support. ... more ILE has new functionality that we, as the Co-Editors of this journal, are asking you to support. It is called collections, and it provides a permanent record of published articles in the journal themed together and easily accessible from the collections page and from each article in the collection. Even if you have not published in the journal, but refer to several themed articles repeatedly, please consider creating a collection page of articles for you and others to use repeatedly.
Http Dx Doi Org 10 1080 10494820 2012 758828, Jan 23, 2013
By and large teachers know how to motivate students by lecturing and interacting with them by sta... more By and large teachers know how to motivate students by lecturing and interacting with them by standing in front of them in a classroom. How can a teacher compete with words and gestures against the dramatic power of carefully crafted interactive computer learning environments? This is a question that has troubled me for years, as I watched teachers trying out dramatic new technologies from intelligent tutors to video games in their class on a single shot trial basis, watching their students become animated and excited by the new ...
Intelligent Tutoring Systems: Lessons Learned, May 1, 1988
KNOWLEDGE ACQUISITION 17 that technicians love playing these debugging games with each other, and... more KNOWLEDGE ACQUISITION 17 that technicians love playing these debugging games with each other, and that novices find them compelling learning vehicles. The connections between this knowledge-acquisition method and reciprocal teaching (Palincsar & Brown, 1984) are obvious and suggest that games designed in this way may provide superior forms of instruction. The significant benefit that these procedures have for resolving the knowledge-acquisition bottleneck is that they make it possible for relatively novice knowledge ...
By and large teachers know how to motivate students by lecturing and interacting with them by sta... more By and large teachers know how to motivate students by lecturing and interacting with them by standing in front of them in a classroom. How can a teacher compete with words and gestures against the dramatic power of carefully crafted interactive computer learning environments? This is a question that has troubled me for years, as I watched teachers trying out dramatic new technologies from intelligent tutors to video games in their class on a single shot trial basis, watching their students become animated and excited by the new ...
Intelligent tutoring systems: Lessons learned, May 1, 1988
The MACH-III system is intended to be a training device in which state-of-the-art techniques in a... more The MACH-III system is intended to be a training device in which state-of-the-art techniques in artificial intelligence and cognitive science are applied to support the training of novices in troubleshooting complex electronic devices at the organizational maintenance level. Organizational maintenance encompasses those maintenance and repair activities that are performed on equipment in the field, that is, at the tactical site to which it is deployed, to render it operational as quickly as possible. Organizational maintenance generally ...
1. Stimulated by Daniel Goleman's bestseller, the concept of Emotional Intelligence (EI) has... more 1. Stimulated by Daniel Goleman's bestseller, the concept of Emotional Intelligence (EI) has become enormously popular in recent years. Originally formulated by Peter Salovey and John Mayer in 1990, three major components of EI were postulated: appraisal and expression of emotion, regulation of emotions, and utilization of emotions (with further subdivisions of each of these branches). Seven years later these authors presented a modified version of EI and the first performance test (ie, Multifactor Emotional Intelligence ...
These memory experiments follow Psotka et al. (2022 b) with distorted binary patterns and examine... more These memory experiments follow Psotka et al. (2022 b) with distorted binary patterns and examine how chunking improves memory for binary patterns, and how many repetitions are needed for the first correct prediction. A simple paradigm is used, where less than 1 element of each base binary pattern of length 16 is randomly complemented, on average, and presented in a list sequences, presented discretely and sequentially. Each sequence is presented one element at a time and the next element is guessed. After each presentation of all 16 elements (onre presentation) the Experimenter/Subject guessed how many distortions were in the pattern Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication. On each trial, the S/E chunked the sequences into subsequences of length 4. This conscious rule arose after many trials and was tested in a separate experiment. Chunking dramatically increased the learning rate but learning was still a function of the redundancies in the sequence.
These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and a... more These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and add unconscious creation of prototypes as measured by the Syntely algorithm. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. Here we examine the number of changes or distortions in binary patterns of length 16. A simple paradigm is used, where about 1.5 to 4 elements of a binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What is the learning rate as measured by the Syntely algorithm (correlations between predicted elements versus test elements of the base pattern for sequences where the average number of changes vary)? This experiment then focuses on the easier patterns, recognizable consciously within about 200 trials to look at the Syntely algorithm's rate of learning. Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and a... more These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and add conscious verification by the experimenter and learning rate over 100 trials. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. Here we examine the number of changes or distortions in binary patterns of length 16. A simple paradigm is used, where about three to 6 elements of a binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What is the learning rate (correlations between predicted number of changes and actual changes)? Previous experiments show that the base pattern is selected way above chance with 4 to 16 elements. This experiment focuses on the easier patterns, recognizable consciously within about 200 trials to look at the rate of learning. Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
In an annual review of memory Kahana (2020) argued that ", our core theoretical understanding of ... more In an annual review of memory Kahana (2020) argued that ", our core theoretical understanding of human memory is still firmly grounded in the list recall method pioneered by Ebbinghaus and Müller in the late nineteenth century". Perhaps it is time to change the focus to inductive memory. Ebbinghaus also conducted other experiments on association memory in which he proposed and found evidence that association memory is essentially unlimited, although weaker with distance. This may now be seen as the basis for convolution and distributed memory, which in turn is the basis for factor analysis, regression, and modern artificial intelligence computations. Our experiments are the beginning of a series of monographs exploring binary sequences to confirm and extend Ebbinghaus' findings to inductive memory. This paper, first prepared in 1978, argued that another paradigm must be added to the standard view of memory: Inductive memory (Psotka, 1975) as an extension of association and convolution. Binary sequences of length eight were
These prototype abstraction experiments follow Posner and Keele (1970) and Garner (1974) with bin... more These prototype abstraction experiments follow Posner and Keele (1970) and Garner (1974) with binary patterns and add conscious verification by the experimenter. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. A simple paradigm is used, where about three to 7 elements of each binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. What are the strengths and limitations of our ability to discern the base pattern? The task is to select the most representative pattern in the set when it occurs within the set. Previous experiments show that the base pattern is selected way above chance with 4 to 10 elements. This experiment extends the number of elements to 16 and explores the main factor, the mean number of perturbations, that measures the similarity to the base pattern (which is not trained explicitly). Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
Planton et al. (2021) conducted a very thorough set of experiments about memory for binary sequen... more Planton et al. (2021) conducted a very thorough set of experiments about memory for binary sequences, comparing a broad variety of different models. Since memory is fundamental to prediction, I decided to examine their first experiment with 16 elements in detail. Besides, they were one of the very few to cite my work (even though the citation was a little wrong), Psotka (1975). So, I decided to replicate their Experiment 1 to see how much of their memory results could be explained by the prediction components. The Syntely measure I use is a kind of association, autocorrelation, or convolution, with the most recent items used as a recall cue. It also uses a kind of chunking of runs at the terminal part of the sequence when used as the recall cue (cf. Psotka, 1975). Since their paper focuses on the power of recursion over association, I thought this would be a useful comparison. Fortunately, they use a kind of prediction paradigm with reaction time (and error rates) to predict tones after the sequences have been learned, and this fits the Syntely paradigm perfectly. They acknowledge that Experiment 1 is the most informative experiment in this paper because: "… we predicted that the effect of LoT complexity in the proposed language of thought would increase as the sequence gets longer." Cf. p. 43
The After Action Review (AAR) process provides a powerful methodology that in the context of trai... more The After Action Review (AAR) process provides a powerful methodology that in the context of training maximizes the benefits of exercises by enabling a unit to learn from experience by systematically reflecting on their strengths and weaknesses. We have developed a tool that supports the AAR process, essentially extending an Observer Controller's (O/C) reach automatically. This tool was developed with two training contexts in mind: live STX lane convoy training at the National Training Center (NTC) and simulated convoy training using DARWARS Ambush! at the Mission Support Training Facility at Fort Lewis. At NTC, live radio communication is captured during training, while with Ambush! communication using voice over IP (VOIP) is recorded. The tool automatically converts recorded speech to text and then analyzes the text, using advanced statistical machine learning technologies, to determine a unit's performance and identify critical incidents, leading indicators, and other t...
ILE has new functionality that we, as the Co-Editors of this journal, are asking you to support. ... more ILE has new functionality that we, as the Co-Editors of this journal, are asking you to support. It is called collections, and it provides a permanent record of published articles in the journal themed together and easily accessible from the collections page and from each article in the collection. Even if you have not published in the journal, but refer to several themed articles repeatedly, please consider creating a collection page of articles for you and others to use repeatedly.
Http Dx Doi Org 10 1080 10494820 2012 758828, Jan 23, 2013
By and large teachers know how to motivate students by lecturing and interacting with them by sta... more By and large teachers know how to motivate students by lecturing and interacting with them by standing in front of them in a classroom. How can a teacher compete with words and gestures against the dramatic power of carefully crafted interactive computer learning environments? This is a question that has troubled me for years, as I watched teachers trying out dramatic new technologies from intelligent tutors to video games in their class on a single shot trial basis, watching their students become animated and excited by the new ...
Intelligent Tutoring Systems: Lessons Learned, May 1, 1988
KNOWLEDGE ACQUISITION 17 that technicians love playing these debugging games with each other, and... more KNOWLEDGE ACQUISITION 17 that technicians love playing these debugging games with each other, and that novices find them compelling learning vehicles. The connections between this knowledge-acquisition method and reciprocal teaching (Palincsar & Brown, 1984) are obvious and suggest that games designed in this way may provide superior forms of instruction. The significant benefit that these procedures have for resolving the knowledge-acquisition bottleneck is that they make it possible for relatively novice knowledge ...
By and large teachers know how to motivate students by lecturing and interacting with them by sta... more By and large teachers know how to motivate students by lecturing and interacting with them by standing in front of them in a classroom. How can a teacher compete with words and gestures against the dramatic power of carefully crafted interactive computer learning environments? This is a question that has troubled me for years, as I watched teachers trying out dramatic new technologies from intelligent tutors to video games in their class on a single shot trial basis, watching their students become animated and excited by the new ...
Intelligent tutoring systems: Lessons learned, May 1, 1988
The MACH-III system is intended to be a training device in which state-of-the-art techniques in a... more The MACH-III system is intended to be a training device in which state-of-the-art techniques in artificial intelligence and cognitive science are applied to support the training of novices in troubleshooting complex electronic devices at the organizational maintenance level. Organizational maintenance encompasses those maintenance and repair activities that are performed on equipment in the field, that is, at the tactical site to which it is deployed, to render it operational as quickly as possible. Organizational maintenance generally ...
1. Stimulated by Daniel Goleman's bestseller, the concept of Emotional Intelligence (EI) has... more 1. Stimulated by Daniel Goleman's bestseller, the concept of Emotional Intelligence (EI) has become enormously popular in recent years. Originally formulated by Peter Salovey and John Mayer in 1990, three major components of EI were postulated: appraisal and expression of emotion, regulation of emotions, and utilization of emotions (with further subdivisions of each of these branches). Seven years later these authors presented a modified version of EI and the first performance test (ie, Multifactor Emotional Intelligence ...
These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and a... more These prototype abstraction experiments follow Posner and Keele (1970) with binary patterns and add conscious verification by the experimenter. Conscious iconic representations produced by unconscious convolution of binary patterns of multiple lengths of 16 are influenced by many factors. A simple paradigm is used, where about three to 7 elements of a binary pattern of length 16 are randomly complemented and presented in a list of 100 sequences. The mean number of elements complemented in each variant is varied from about 3 to 7, increasing the difficulty of recognizing the base sequence. What are the strengths and limitations of our ability to discern the base pattern? The task is to select the most representative pattern. Previous experiments show that the base pattern is selected way above chance with 4 to 10 elements. This experiment extends the number of elements to 16 and explores the main factor, the mean number of perturbations, that measures the similarity to the base pattern (which is not trained explicitly). Only one Subject (the Experimenter S/E) was used but the results are so clear they demand communication.
Uploads
Papers by Joe Psotka