Measures:

• Recall
• True / False
• Perform task

Subjects:

• 85 students
• 17 per condition
• Balanced with respect to familiarity with the task

Apparatus:

• Digital clock
• Spool device

Tasks:

• How to use a timer clock
• How to make and use a spool device

Procedures:

• Read text once
• Actions dependent on condition (see device, etc.)
• Recall instructions
• Answer True / false questions
• Perform task with use of procedural text

Within subjects analysis:

• Kind of task / Kind of procedural text

Between subject analysis:

• Text order (clock-spool versus spool-clock)
• Condition (read-do versus read-see versus read)

Slide:

Predictions

Text representation theory

Task performance:

• Good situational models are best for task performance
• Read only leads to worse performance, because available information is limited to the given text (worse situational representation)
• Read-imagine provides subject with incorrect situational model

1. Read-do
2. Read-see do
3. Read-see
4. Read only / Read-imagine

True / False:

• Read only, read-see, read-imagine provide subject with well developed text bases (propositional representations)
• With read only, read-see, read-imagine there is no interference, subjects can focus better on the text

1. Read-see / Read only
2. Read-see do / Read imagine
3. Read-do

* Recall:

• Read only, read-see, read-imagine provide subject with well developed text bases (propositional representations)
• With read only, read-see, read-imagine there is no interference, subjects can focus better on the text

1. Read-see / Read only
2. Read-see do / Read-imagine
3. Read do

Dual Coding theory

Task performance, True / False, * Recall:

• Read-do contains more codes (also proprioceptive) than sequentially read-see-do, read-see, read-imagine, read only

1. Read-do
2. Read-see do
3. Read-see
4. Read-imagine
5. Read only

Slide:

Results

Main effects:

• Clock better than spool over all
• Determine own reading time worse than fixed reading time with respect to recall

Task performance:

1. Read-see do
2. Read-do / Read-see
3. Read-imagine / Read only

True / False:

1. Read-see
2. Read only
3. Read-see do / Read imagine / Read-do

Recall:

1. Read-see / Read only
2. Read-see do / Read-imagine
3. Read do

Conclusions

• Access to device enables a better development of a situational representation which leads to a better task performance
• Task orientation leads to it that the reader focuses less on the text , which leads to a poorer propositional representation, which leads to poorer recall
• When textbase is well established a good situational representation can enhance memory for the text, which leads to recall and True / false performance (of read-see)
• Doing does lead to a (near to) equal situational representation, and thus to an ‘equal’ task performance as read-see
• Doing leads to a worse situational representation, and thus worse task performance than read-see-do

• Do you teach what you want to teach? For example, if you teach a student with by means of ‘Read-do’, he/she will be able to perform fairly well, but will not perform very well on a true / false or recall test. So, if you want a student to obtain much ‘knowledge’ about a system, it is not a good decision to teach him/her by means of ‘read-do’.

• Do you test what you want to test? For example, if you teach a student by means of ‘reading-see-do’, he/she will be able to perform the task well, but will not perform very well on a true / false or recall test. So if you test such a student with such tests, for example with CBT, you will not get a good impression of his/her skills.

• Mimicking better than watching the experimenter or doing it by yourself? (Tomassello, children learn by mimicking, natural way of learning for human)

• Effect of sequence? (First read, creating good text base, then watch it do, then doing it yourself, etc)

• All on screen? (Seeing of animations, watching good procedures on screen, doing with VR objects on screen, etc)

• What design when goal skill-, rule-, or knowledge-based behavior? For example, does dual-coding theory fit within different circumstances, such as learning skill-based behavior?

Slide:

Onderzoeksproject Universiteit Twente en Holland Signaal
Mona Claessens
27 September 1999

1. Activiteiten afgelopen jaar

2. Toekomstplannen

Slide:

1. Activiteiten afgelopen jaar:

Universiteit Twente	Holland Signaal
Theoretische onderbouwing voor de hypothese dat onder bepaalde omstandigheden het parallel aanbieden van informatie gunstig is.	Praktische toepassing van ‘Computer Based Training’ en ‘Virtual Reality’ bij training van ‘Signaal-klanten’.
Theorieën moeten leiden tot een geschikt design van ICT-applicaties die gebruikt worden bij het aanleren van verschillende taken.
Literatuurstudie: Taken Informatieverwerkingstheorieën Virtual Reality Bevindingen: Indeling taken Bepaling design Definiëring Virtual Reality	Veldonderzoek: Demonstraties Volgen cursus Gesprekken / Interviews Bevindingen: Definiëring doelgroep Definiëring context Leidt tot specifieke design eisen

Slide:

2. Toekomstplannen

Het vinden van de ideale ‘taak-design-match’:

1. Het vinden van prototypische ‘Holland Signaal’ taken bij elk taak niveau, zoals gevonden in de literatuur.

    

2. Het ontwikkelen van Computer Based Training software om deze taken te leren.

    

3. Het testen van verschillende designs voor het leren van deze taken, gebaseerd op informatieverwerkingstheorieën.

    

4. Het vergelijken van de taakprestatie met de tijd die er voor nodig was de taak te leren.

Chapter D.2: Course

There were 7 participants of the course. They were all Indonesian maintainers. Three of them were from a shipyard (civil) and four of them were from the marine (military).

The first day students were explained the necessary theory (from 10.50h - 12.50h and from 13.20h - 14.08h) and were made familiar with the apparatus (14.25h - 15.50h).

The second day the students got the opportunity top practice more with the apparatus and meanwhile some more theory was provided on the basis of the practicing.

The teacher used a blackboard and sheets that showed schemes and print outs from display-screens from the apparatus. He explained things very slowly and used examples, anecdotes or even jokes. He evoked participation by asking questions before giving the information. After the explanation of important information he was quit for short period, so students could write things down. Regularly he asked whether they understood things. When sheets were shown he referred to the pages of their manuals, were the figures etc of the sheets were also available.

There was not very much response, maybe because of the language problem. Some students could nearly not speak English.

The students came in at about 9.15u. They were asked to study the accompanying manuals. At 10.50u the lesson began with a question about an excursion they had made the day before. After a short 'chat' an introduction was given to the things that would be taught. Then it was told what kind of lesson materials (manuals) they had to use for this course.

The course was divided in two theoretical parts: theory about the MOC (Multi Operator Console) and theory about the GSS (x). The MOC is a console that is mainly used by operators and the GSS is a console that is specifically used by maintainers. They also have the opportunity to use a PIOD (x), but this is not explained in this course.

The main task of this maintainer course is to find check the state of the system and to find faults. Comparing two windows (on the MOC or on the GSS) available at the same time can be done for checking the system. In these windows contain icons of the subsystems involved (for example radar's, guns etc.). These icons are marked by colors that refer to the states they are in (green = available, orange = degraded, red = not available and grey = not connected). By comparing these icons the states of the subsystems can be determined.

Faults can be found in four different ways. They can be found by clicking on the icons from the subsystems that they want to check for faults. Another window appears that shows where in the system there are errors. From several aspects of the system it is showed if they are available or not. When one of the aspects are not available it there is a problem. Another way to spot faults is by looking at the statistics window, here there are graphs that show the load of the system. The third way is to check the system with the help of the BITE (Built in Test) of the system. This is an automatic check of the system for occurring faults. Via the PIOD these test can be seen remotely. Finally, a maintainer could walk to the system itself to check for defects.

For the practical part of the course the group was divided into two subgroups, three people were seated behind the GSS and four people were to work with the MOC. They were asked to fulfil several assignments, but it did not happen in a very structured way. The lesson was mostly meant to play with and get familiar with the systems. When a problem occurred or a question was asked the teacher asked a general question about this and they had to find out the answer themselves.

The second day they got the opportunity to practice more and got more theory on the basis of practical rehearsal.

This was a course in the highest level of faultfinding. This means that it concerns finding in what subsystems there are faults. Deeper faultfinding concerns were in the subsystems there are faults.

Chapter D.3: Summer school

Course 1

Human thinking: Boicho Kokinov

Short description contents

This course will focus both on classical topics in psychology of reasoning such as deductive and analogical reasoning and on topics relating human thinking with development, perception, memory, and natural environment. The problems studied will range from simple syllogistic tasks to scientific discoveries and the methodology used will range from laboratory experiments and the natural observation to computer simulation.

Summary relevant information

According to Gentner (1989) an analogy making is mapping knowledge from one domain (the base) into another (the target).

Sub processes of analogy making:

• building a representation
• selecting a source (retrieval = mapping properties)
• mapping (even happens in simple recognition processes)
• transfer (deductive reasoning)
• evaluate
• learn

Analogies can be of properties, but also about relationships. Analogies are much used for scientific reasoning (Dunbar). Researchers only use structural analogies. They use analogies for formulating, designing and explaining.

Gentner has a structure mapping theory. This mapping model describes the transfer of systems of relations.

• discard distributes
• preserve relations
• choose systems of relations

There is a difference between analogical representation and analogical reasoning. With analogical reasoning you need to transfer rules.

The ability to analogical reasoning starts at a very early age, deductive reasoning starts later. Infants use imitations, later they recognize when someone is imitating him/her. They only imitate active exposures.

Context and priming effects develop with time. Priming is the activation of already processed information. A characteristic of priming is that it does not involve explicit remembering. Experiments were done with ‘priming problems’ (P) and ‘target problems’ (T). First priming problems are given to the subjects and after a certain amount of time (delay) a second problem is to be solved. It is measured whether certain concepts from the priming problem are activated during solving the target problem. Thus, it is investigated whether priming problems influence target problems.

We can find priming with problem solving but it disappears with time.

Priming:

Time

P T

Far priming:

Time

Very far priming:

Time

Priming Effects:

Time delay

Kokinov found context effects on problem solving. He presented subjects a target problem on a paper with an "irrelevant" context problem. They were asked to solve only the target problem. Both problems were accompanied by an illustration. Results showed that answers to the target problem differed according to the different kinds of context problems. It seemed to be that the illustrations were perceived simultaneously and that the context activated certain concepts.

Course 2

Perception-based spatial reasoning: Reinhart Moratz

Short description contents

The integration of perception and the action with processes of reasoning and cognition creates a foundation for intelligent behavior. Aspects of space are central to both perception and the action. In the field of artificial intelligence great success was achieved in dealing with formal knowledge and also with sensor data alone. But to bridge the gap between symbolic and sensor data remains difficult. Spatial reasoning is a field in which the relation between different modalities of knowledge can be studied in a prototypical way.

Summary relevant information

Human visual system:

Pre-attentive processes

• parallel and fast over the whole visual field
• segmentation and grouping
• localization of objects

Attentive processes

• object identification
• reconstruction of spatial parts
• scene analysis

Feature integration theory (Treismann)

When an object can be located by attending only to one feature this search can be done in parallel

When an object can only be located by attending to conjunctions of features this search is done in a serial manner.

Context is important. The scene context effect shows that when objects are shown in unusual contexts, such as a bench in the city they are easier to trace.

Course 3

Models of memory: Richard M. Shiffrin

Short description contents

The course provides analysis of contemporary memory models, though intensive examination of models of each type, and evaluation of their ability to handle the critical data in the field. The models considered are designed to apply to implicit and explicit memory tasks, recognition and recall procedures, free response and signal-to-respond data, and paradigms involving priming and cueing. Models for both accuracy and response time will be discussed. Both computer simulation and mathematical models will be covered.

Summary relevant information

Types of memory:

• Explicit or episodic memory is build up by personally identifiable events that are contextually defined.
• Generic memory is build up by general knowledge, knowledge that is not usually personal or identified by specific contexts.
• Implicit memory is build up by knowledge that is already present in Long Term Memory

Short Term Store divides experience into episodes. These episodes are remembered for a while and then forgotten. This is what they cal retention. The longer the retention intervals the closer to the Long Term Store.

When an event occurs also the content information is stored. Repetitions of the same item in a similar context will often lead to both storage of a new episodic image and the storage of a previous episodic image. If the context of the item varies, you add possible contexts. A group context is created separated from the specific contexts and personal references, general knowledge.

List length effect is the phenomenon that the more items are in a list to be remembered, the less good items can be remembered.

To recall information from memory a sequential search is needed. Items are sampled, searched and checked. The process of recall is slow and not as easy successful as recognition.

To recognize earlier obtained information, there is only one process is needed that works in parallel. Recognition is a fast process with more chance on success than recall.

Recognition is easier on the basis of additional contexts. This is because there are more possible matches with memories of earlier events.

With identification implicit memory is used and with recall explicit memory is used. A difficult task must be done on the basis of implicit memory.

Episodic semantic trace does not chance much over time due to delay. Context does.

It is very difficult to interpret results with response times. There is the complicated relation between accuracy and speed. Speed increases at the expense of accuracy. To deal with this problem the items must be studied longer.

100% % Correct 95%
	Fast                         Slow Response times

Own consideration: Parallel can be simultaneous in terms of receiving the information and in terms of processing the information. Pictures can be processed in a parallel manner (see Treismann’s theory of feature integration), but text always has to be processed in a serial manner.

Course 4

Categorization: Evan Heit

Short description contents

This course addresses to categorization and interrelated cognitive abilities such as recognition memory and inductive reasoning. Topics to be covered include models of categorization, how category learning is constrained by background knowledge, whether there is more than one system of categorization, and how categories are used for inductive reasoning.

Summary relevant information

Categorization is closely linked to other cognitive mechanisms such as memory judgement, problem solving, language acquisition, semantic similarity etc..

With categorization there are two processes involved:

• Integration processes - Previous knowledge provides starting up a representation and revises as new category members are observed (Bayes).
• Weighting processes – Previous knowledge has a weighting form effect. Some category members have greater impact on learning than others. Selective weighting involves strategies (resource allocation) and memory (some members are stronger anchored in memory)

Fast categorization:

• People start with initial category representation based on prior knowledge, equivalent to some number of prior examples
• This representation is gradually revised as new observations are integrated into the representations

Slow categorization:

• Looking for incongruity

Inductive reasoning = Probabilistic reasoning involving projection of properties of some event or category to another similar item.

Inductive reasoning is from general to specific:

All Americans are friendly
Evan is an American
_____________________
Evan is friendly

Deductive reasoning is from specific to general (beyond information you’ve got):

American 1 is friendly
American 1 is friendly
___________________
All Americans are friendly

Similarity promotes inductive reasoning. Bat looks like an owl, but is more similar to a stork.

Course 5

Child language development: Michael Tomasello

Short description contents

This course focuses on basic processes in first language acquisition from the point of view of children's broader skills of cognition, communication, and symbolization. Topics include: cognitive and functional linguistics, pragmatics and word learning., early syntax, individual and cross-linguistic variation, and theories of language development.

Summary relevant information

Children learn by participating in a context, not by just pointing out external things. Children are very good in associating, but that is not the same as learning a word or learning language. You have to make them understand the function.

Processes of language

• Cultural learning (imitative learning); reproducing linguistic forms and meaning
• Analogy / schematization (making abstractions)
• Structure combining (making novel combinations)

Course 6

Short description contents

This course is about the debate against the notion that language is innate.

Summary relevant information

None. Her vision is that language is not an individual thing that is localized in a particular part of the brain. It is learned and distributed across the brain.

Course 7

Culture and Cognition: Roy D’ Andrade

Short description contents

This course is the study of the way in which cultural understandings are psychologically organized and used. The development of a theoretical framework to understand this problem began with work on semantic features and taxonomies and has continued to current work on schemas, neural nets, and cultural models. This course will cover basic findings of this field and also involve students in a series of exercises using ethnographic interviewing to investigate cultural models concerning honor and violence. Cognitive anthropology has connections to standard courses in cognitive psychology, especially work on categorization, reasoning, and memory. It also has relations to work in linguistics on semantics and to work in artificial intelligence.

Summary relevant information

None. Roy D’Andrade has developed a FOLK model that explains human behavior.

Stimuli --> [Perception --> Thought --> Whish --> Intention ] --> Response

Course 8

Short description contents

This course will introduce the participants to what I believe to be one of the main challenges for future cognitive science: How to reconcile Emergence and cognition, with special attention to the social domain. The course will take into account and compare literature from AI (in particular agents and MAS), cognitive science (social psychology), social sciences.

Summary relevant information

None. He showed models about social interaction between agents.

Conlusies

Behandelde onderwerpen

• Heeft me op de hoogte gehouden van wat er gaande is in de cognitieve wetenschap
• Heeft me nuttige informatie / ideeën geleverd voor in mijn proefschrift (zie relevant summary’s).

Kennismaking met verschillende manieren van lesgeven

• Student centered versus content centered approach
• Eigen onderzoek propageren versus gedegen beeld geven van recente ontwikkelingen inclusief voors en tegens

Kennismaking met verschillende manieren van onderzoek doen

• Geïsoleerd werken aan eigen model/ theorie versus samenwerken met onderzoekers uit eigen discipline / kruisbestuiving
• Explorerend versus traditioneel, vasthouden / trouw zijn aan (achterhaalde) concepten

Kennismaking met andere studenten, onderzoekers, universitair (hoofd)docenten.

• Kennisgemaakt met mensen uit Bulgarije, Hongarije, Rusland, Roemenie, Servië, Polen, Italië, Duitsland, Amerika en Australië.

Uiteindelijk ben ik naar al de ‘lectures’ geweest, behalve naar die over ‘visual perception’ (tijdens de eerste les bleek alles oud nieuws te zijn voor mij, bovendien had ik moeite met de manier van lesgeven). Het was erg jammer dat Prof. John Hayes (Applying cognitive science to instruction) niet in de gelegenheid was te komen (oog operatie).

Chapter D.4: Project (afstudeerproject)

1. AVP betreffende computer based training

Er is een aanvraag proefneming gedaan door de subafdeling Training van de afdeling ILS te Holland Signaal. Met deze proefaanvraag wordt toestemming gevraagd om nader onderzoek te mogen doen naar het volgende probleem: In hoeverre zijn Computer Based Training producten (toegespitst op Signaal producten), haalbaar, wenselijk en toepasbaar. Het onderzoek zal gedaan worden aan de hand van een concreet signaal product, het Principal Work Station (PWS), zoals dat voor het LCF-fregat ontwikkeld wordt.

Ik heb de doelstellingen helpen formuleren en daar kwam het volgende uit:

Met deze AVP wordt gestreefd om kennis te vergaren over de volgende zaken:

• Is CBT nodig om de concurrentiepositie te versterken / bij te houden?
• Is CBT belangrijk voor het kwaliteitsimago van een bedrijf?
• Is CBT nodig om te kunnen blijven voldoen aan de toenemende ‘veeleisendheid’ van de klant (steeds kortere cursussen zonder kwaliteitsverlies)?
• Blijkt uit de kosten en baten analyse dat het mogelijk, wenselijk en slim is CBT bij ‘Holland Signaal’ in te voeren?
• Is het winstgevend om CBT als op zichzelf staand product af te zetten op de markt (is er interesse voor dit soort trainingsprodukten)?

Kort samengevat: ‘Moeten’ wij het, kunnen wij het en willen wij het?

Het project zal de volgende samenstelling hebben:

• Deskundige m.b.t. trainingsapplicaties en ervaren met Signaal werkwijzen (Ton Muller; Holland Signaal, ILS Training)
• Deskundige m.b.t. het bouwen van (CBT) applicaties (Derk Groeneveld; Holland Signaal, ILS Training)
• Deskundige m.b.t. het ergonomisch ontwerp en de didactische opbouw van (CBT) applicaties (Mona Claessens; Universiteit Twente, Toegepaste Onderwijskunde)
• Deskundigen ter ondersteuning m.b.t. onder andere de consultancy (Hans Clerx, Han Wientjes, Ronald Boers, Harry Menninga etc.; Holland Signaal, respectievelijk ILS Training, RD & T, ?? en ??)

Er wordt ook nagedacht over het betrekken van een afstudeerstagiaire van Toegepaste Onderwijskunde.

Achtergrond
Hollandse Signaal apparaten B.V. is een defensie organisatie die radar apparatuur ontwikkelt en produceert voor de marine. De afdeling ‘Integrated Logistic Support’ zorgt voor de ondersteunende producten behorende bij de Signaal Systemen, zoals reserve onderdelen, speciaal meetapparatuur, gereedschappen, (digitale) ‘manuals’ en opleidingen. Het onderdeel training verzorgt de opleidingen voor de klant met betrekking tot de bediening en het onderhoud van ‘Signaal apparatuur’. Bij de afdeling training is behoefte aan een onderzoek naar de het nut van computer gestuurd onderwijs voor klanten opleidingen.

Opdracht

• Inventarisatie van op dit moment in de markt beschikbare ‘CBT-toolings’ die voldoen aan de eisen van het ‘International Training Centre’.
• Zoeken naar nieuwe toepassingen / uitbreidingen van CBT om reeds gestelde eisen aan te vullen.
• Mogelijkheden om met minimale inspanning bestaande tekeningen en andere informatie uit gebruikte CAD / CAM systemen over te kunnen hevelen naar het CBT pakket.
• Het bouwen van een didactisch verantwoord CBT prototype.

Vereisten:

• Interesse in de praktijksituatie bij Holland Signaal
• Ondernemend
• Zelfstandig

Voor meer informatie kun je contact opnemen met één van de begeleiders:

Mona Claessens: Claessens@edte.utwente.nl; 053-4893873
Hans Clerx: db946@signaal.nl; 074-2482504
Rik Min: Min@edte.utwente.nl 053-4893585
Jef Moonen: Moonen@edte.utwente.nl