Response to an evaluationThe following concerns the examiners' reports on the Honours thesis "A Multi-level Context Evaluation System". The thesis was submitted in partial fulfillment of the requirements of the degree of Bachelor of Information Technology (Honours) in October 1999, at Griffith University, Brisbane, Australia. It should be pointed out the thesis does not represent a forerunner of the Otoom model.
All in all it represents an illustration of purposefully contrived criticisms in the service of a higher agenda.
The referee's report was not much better. For example, it made this interesting observation: "Overall, the writing quality and organisation of the thesis is very poor. There are many spelling and grammatical errors, and themes and arguments are virtually absent. There is little reference to the peer-reviewed scientific literature."
Here are the stats:
page count: 153;
word count: 46,596;
references: Part 1: 23, Part 2: 11, Appendix/Overview: 20, total: 54;
Spelling mistakes found: 1. tradespeople (trades people); 2. sphisticated (sophisticated); 3. handprinted (hand printed);
Grammatical mistakes(?) found: "which" instead of "that", but mostly in sentences containing a comma.
Note: Just read through Hierarchical Temporal Memory - Concepts, Theory, and Terminology by Jeff Hawkins and Dileep George, Numenta Inc. Their HTM system (the company was founded in 2005) uses a "conditional probability table", a "matrix of numbers where the columns of the matrix correspond to the individual beliefs from one node and the rows correspond to the individual beliefs from the other node". This sounds suspiciously like the evaluation matrices used in the MCES (see thesis, 2.5. Context Identification and Evaluation, p. 69). Did the examiners have any idea what they were reading??
Note: After the honours year at Griffith the plan was to move to the University of Queensland, St. Lucia campus, to continue under the auspices of a PhD course. The book "On the origin of Mind" was intended as the thesis for the doctorate. Towards the end of 1999 I had already reserved a place under Dr Simon Dennis, so much so that the offer was held open during much of the appeal process. As the appeal dragged on I had to let it go. The University of Queensland's entry standards had always been high. In The Good Universities Guide 2008 UQ achieved a top five-star rating for research intensity and toughness to get a place at its St. Lucia Campus. If my academic state of mind had been so wayward, would I have got the offer and for that length of time?
This then is what I had to say in the appeal:
Chair of the Assessment Board
Examination and Timetabling Centre
Griffith University 4111
Re: Appeal against the grade awarded for my Honours thesis, "A Multi-level Context Evaluation System (MCES)"
In this appeal I will in particular refer to the two examiners' reports (forwarded to me by the Honours Administrative Officer, CIT) and their specific criticisms, but will also make more general comments based on the overall tone of those two reports. The two reports are titled REPORT ON MARTIN WURZINGER'S HONOURS THESIS (from now on referred to as Report 1), and Observations and Comments on a Document Submitted by Martin Wurzinger for the Written Component of an Honours Dissertation (from now on referred to as Report 2).
This is not intended as a further explanation of the thesis material itself - in my opinion the document contains descriptions and references for someone to be sufficiently informed about the topic, provided the references are used to supplement any already existing knowledge base a reader ought to have. There is no denying that the areas of cognitive science and/or artificial intelligence (AI) are highly specialised; without some prior knowledge, particularly about the issues raised in the thesis, it would be rather difficult for anyone to come up with an informed opinion regarding its contents.
I will conclude this appeal by providing a brief reference to the history of this undertaking, not because I feel it should form part of the assessment per se, but because of the doubts cast upon the background of the thesis and to show that the material is not "unorganised and rather self-indulgent" as one of the examiners put it.
Whenever a reference to a page number is being made without further description, that page refers to a location in the thesis; otherwise the name of the text is mentioned alongside.
To begin with may I make the following observations:
Any description or presentation of an idea, a model, and its implementation, ought to provide the recipient with a certain background information as to why such an idea, model etc, is found to be important. Especially if the subject matter is controversial and subject to possible criticism, an introduction to the general problem, its significant features, and the reasons why this or that approach had been taken, ought to be supplied with the product (in this case the thesis) itself. Once the preliminary information has been offered, it remains then to describe the technical details of the implementation, so that at any given point reference can be made to, in this case the computer program, and the entire process be confirmed on a step-by-step basis.
The failure to do just that is of course the basis of the two examiners' assessment, but through the following pages I hope to show that the criticism is unjustified.
Response to Report 1
In para. 2 of the report the examiner states that it "contains sweeping and unjustified generalisations that vacillate between philosophy and psychology, the neurosciences, and sociology and political theory".
I do not deny for a moment that in the thesis I have not confined myself to one single stream of science. It is widely recognised by those involved in the field of AI that here is one area which does not allow itself to be categorised by a single set of features, and that its multi-faceted themes can, in fact ought to be, approached from a great variety of directions. On p. 8 I cite Mira and Delgado, who emphasise the need for an interdisciplinary approach.
The reason why one or the other topic has come to be dealt with by a particular field of science has to do with the topic content itself, but also can be due to historical reasons. Not all fields of science have advanced in tandem over the last few hundred years, and what has been discovered in one area may have remained dormant until rediscovered in a different context by researchers in another. This gives rise to possible fractionising, but also to divergent nomenclature despite the fact that one topic's 'natural' domain may be the same - it just had been subjected to differing foci.
When it comes to human cognition, not only are the possible directions of approach vastly at variance with each other, the history of the interest shown in it goes back literally thousands of years and spans the globe. Whether such interest came from the ancient Jain culture in Southern India in the form of the Pañcástikayására, or from the analysis of human nature by Machiavelli in the European 15/16th century, or whether it is being represented by the myriad contemporary approaches by thinkers anywhere in the world, to claim that its application is only acceptable if constrained to a narrow band of criteria would immediately consign the result to potential failure.
The examiner is quite correct in identifying the many fields (eg philosophy, psychology, neurosciences, etc; para. 2), and it is precisely such a vast arena which needs to be utilised if one wants to make progress with regards to the many questions that human cognition poses.
I do not claim to be an expert in all these fields; my interest lies in the form and content of thoughts as they are created by human minds, and how it is that they can be so diverse (giving rise to varying opinions, values, morals, and their sociological and political equivalents). I have studied these thoughts, or thought structures, in many diverse areas of the world, among different cultures and demographics, a fact I alluded to on p. 9 and onwards. I realise that such a reference is not the traditional listing of an academic paper; but I also know that ultimately the understanding of human behaviour does not only come from an exclusive analysis of controlled experiments and theoretical thinking. As valid as those means are, in the case of human behaviour they can only supplement the insights that can be gained among the people of an African village, the CBD in places like Bombay or London, or the complex social fabric along the river systems of a country like Guyana.
The mentioning of the above listed background is used by the examiner as a means to construct his/her criticism, when in fact it should be seen as a valued characteristic of the thesis (para. 2).
This applies equally to the "30 page appendix of newspaper cuttings from the Courier Mail" (para. 2). The section Queensland - profile of a society to which the author refers constitutes not merely anecdotal incidents gleaned from a newspaper or one's personal experience; the majority of those examples are in fact excerpts from reports, findings, and analyses (referenced by name and date) which have been supplied to build a certain profile of a society, and the aim of this exercise is to show how various assumptions, values, and traditions converge and format the society of which they are a part - just as they do in the case of an individual who is situated within a culture and thinks and therefore acts in a specific manner. Queensland - profile.. has not simply been added ad hoc for readers to find their own reasons or justifications as to its inclusion; the reason is given on p. 25, after a thorough discussion of thought processes and their various levels, to show that as thought structures define, indeed control, one's decision making processes they also influence the dynamics of entire societies and cultures.
Again in para. 2, it is stated that I show how the human mind works "in a series of a priori rules that Martin seems to have made up, because no evidence is provided for them". If "a priori" means that I posited these rules ahead of further explanations, then the epithet seems reasonable. To imply however that they are merely stated for readers to be accepted as gospel is simply not true. Immediately after the listing of the rules I have supplied a detailed description of what I meant, with examples from real life to show that this is indeed how humans think and act. There is no conjecture, presumption, or the need to appeal to readers' credulity; every item comes with a reference to actual situations that are quite common, every-day encounters in any community. The rules and their explanations start at p. 11 and continue until p. 18. "No evidence"? Hardly.
"By page 15 we are into a priori sociology and political science. By page 18 we are into a priori neuroscience." I have already referred to AI as a prime example of overlapping domains. If a certain statement can therefore be assigned the label "sociology" and another that of "political science", then so be it. Human behaviour, as disconcerting as it may seem to some, does not configure itself according to headings and subtitles. If a particular aspect conveniently falls into the area of sociology, and its neighbour has been dealt with under the auspices of neuroscience, does that mean that those two aspects are irrelevant to each other, or worse still, should be denied their significance within the context of an overall model? To assert such an opinion without giving due regard to the nature of the aspects in question seems to be quite "a priori" in itself.
Still within the above context, the report says "the thesis is apparently about sentential or linguistic contexts (p. 115), but Martin tells us that he got his original ideas by looking at social and cultural contexts, and on page 10 he lists the many countries where these observations were made." I fail to see how such a statement could be proffered as a way to downgrade somebody's endeavour, were it not for its subtle wording. Yes, the thesis is about sentential or linguistic contexts - I repeatedly say so throughout the document. At the very next moment however we are given apparent reason to doubt this: "...but Martin tells us that he got his original ideas by looking at social and cultural contexts" (my italics).
In the absence of a thorough understanding of how our mental processes actually work, we can only consider their effects, that is thoughts, their expressions, and how they translate into subsequent action (at this stage of our knowledge EEG's for example do not explicitly describe thoughts, they merely hint at their presence). And the way we learn about our thoughts and their expressions etc, in fact the only way, is through the sentences we form in order to articulate whatever it is we are thinking. So of course the phenomenon of language, its various formats, and what actually is or can be articulated through such a medium, occupies a special place of significance when it comes to learning about human thought processes. Since it is through the saying of sentences that we convey our ideas to each other, and through this process form entire structures of values (which become enshrined in religion, the law, philosophy), the examination of social and cultural contexts seems a good place to start.
Referring to the section about Queensland in the appendix (Queensland - profile of a society), the author mentions my comments after that section where I state that certain conclusions can be drawn from such a profile such as, among others, "A society is only as good as the sum total of its citizens".
In the very next paragraph I explain what I mean by the expression 'sum total' to prevent any possible misunderstandings by somebody who might have a say, mathematical background (p. 144). I am at a loss to understand why the quoting of these sentences could be taken to support the author's undeniable intent at criticism. Certain conclusions can indeed be drawn from such a profile, and I continue to provide a list of these (from p. 144 to p. 146 inclusive). And why am I wrong in saying that a society is only as good as the sum total of its citizens? The report leaves it at that - there are no further comments as to why the conclusions are inappropriate, presumptuous, or just plain wrong.
One could possibly argue that in a thesis essentially about a computer program such descriptions are out of place. The reason for their inclusion however is stated at the very beginning of the appendix (p. 118). By way of preface to the Profile I explicitly state that I have enclosed the entire section for reasons of completeness - I did not want to give rise to the objection that in order to bolster my arguments I have conveniently censored my own work, leaving out certain parts which quite possibly could diminish the arguments' validity (however unfounded such an objection is going to be upon closer examination), especially since the Profile had been forwarded to numerous people over a period of time and the thesis itself is now in the public domain. Moreover, I referred to comments made at the recent World Conference on Science in Budapest that science ought to be seen as an integral part of human society, with all the ramifications such an attitude invokes, including the recognition that AI in this case should have something to say about the human condition.
The report goes on to say "on page 42-44 we have a section on consciousness, which gets the main problem of consciousness back to front, and pages 44-50 are about evolution" (para. 3).
Firstly, there seems to be the implication once again that for a thesis to cover one topic and then another, is some sort of failure of the thesis. I have presented my argument against such a notion in a previous paragraph and so there is no need to repeat myself.
Secondly, in case anyone should doubt the relevance of a topic such as 'consciousness', I have provided a quite recent reference to show that, as philosophical or esoteric as it may seem, the issue of what constitutes that part of ourselves we call consciousness is in fact part and parcel of ongoing arguments in AI. That there is no comprehensive, coherent identification of that phenomenon should be clear to anyone who has ever read any appropriate philosophical treatise at any level of depth. It should be equally clear then that the opinions offered on this subject are very much a consequence of the cultural, ideological, or religious basis from which they originated - exactly my point in terms of human thought structures. I have not in any way criticised any particular attitude in this matter, on whatever grounds, nor have I suggested in any way or form that I myself should be considered an expert on this. I merely provided the not unreasonable suggestion that in the absence of any 'hard facts' it might be a good idea to leave any presumptions aside for the time being, in the context of the thesis. This is a far cry from a demand to deal or not to deal with the topic, nor could it be construed that I have got the problem "back to front".
Still in para. 3, I am also being criticised - certainly implicitly - for dealing with evolution. No further explanation is forthcoming, but if I take the sentence to be written in the general ambience of the report, I am left to conclude that it is inappropriate to deal with any evolutionary aspects in the thesis. Again I must say I find this curious.
To approach an investigation into the human mind, and therefore the human brain, as if such a system was merely the product of some instant, contemporary process, is totally misunderstanding the nature of not only the brain, but organic systems in general. To dwell on that point any further should really not be necessary in a document of this nature, intended as it is for an academic audience.
In para. 4 the author reiterates the idea that the thesis is "apparently about linguistic or sentential contexts", and then goes on to say that my extended notion of 'context' ranges from "the linguistic to the socio-cultural, and sometimes .. the neurological". And so it should, given that ideas are not merely sentences, but also thought processes, not merely world-views, but also activations inside our brains.
Referring to what I called conels, or context elements, the author understands them as semantic primitives. Indeed they can be seen as such, but at the same time the word "semantic" does not only limit itself to linguistics, but applies to anything where the concept of meaning or significance needs to be ascertained in one form or another. This by the way is not my personal interpretation, but can be confirmed by looking up the word in a dictionary. So yes, within the arena of language a context element can be understood as a linguistic entity; at the same time, when talking about neurological activities, we can also use the expression of "context element" to describe a feature which exists at a bio-chemical level. In case anyone should doubt the validity of my choice of words, I have provided several references (on pp. 38, 39, 40) to show that I am not the only one who has looked into that direction. In addition, on p. 39 I have referred back to a previous section (ie 1.2, Levels of Operation) to demonstrate exactly what I mean by conels and how they are to be understood.
The report represents my remarks about conels on pp. 40 and 39 (listed in that order) as if they were mutually contradictory. Considering what I have said above, such a notion can only be sustained if one misunderstands the purpose to which I subjected my choice of words.
In para. 5 the author asks, "is this thesis *just* a computer program, or is it supposed to model the human mind?"
A reasonable question, and one which is answered right there on p. 97; I am not going to repeat it. The author continues (same para.) by citing my opening comments for Part Two (p. 53), and some further comments I made in the Conclusion. The pertinent word in both samples is "functionalities". What this means is that in terms of functionalities, as described throughout the thesis, the program does model the human mind. To explain how has been the reason for going through the various topics presented in Part One, so one may gain an understanding of how one might choose to see the identifiable side of human thoughts; followed by a detailed - and I mean detailed - description of what the program actually does, so that the meaning of contexts, how they feature in either model (human or machine) and how the program handles them can be ascertained by an outsider. The last sentence in para. 5 presents a quote ("..is about the development of a computer program and nothing else") as if I myself had chosen to contradict the idea of encompassing any themes outside the area of computer programming by that remark. That seeming contradiction has been taken from p. 32, where I talk about the practical side of understanding human thought structures and the advantage it confers in certain situations, again in practical, every-day situations. That sentiment has nothing to do with quite different contexts expressed on pp. 53 and 115 respectively.
Para. 6 states that I talk about meaning and content, but that the program does not understand meanings as I myself supposed to have said on another occasion.
The reference to "meaning and content" (page number given is 83) is rather spurious. Neither on p. 83, nor indeed on the directly previous nor the following page is there any allusion to "meaning and content" at all. What p. 83 in fact contains is a brief summary of the program's process so far, so that the reader may enjoy some respite from the quite detailed descriptions so far before continuing with further, equally detailed, explanations. The only time the word "content" appears on this page is in the sentence "To review: we have a matrix which represents the relationship between the current input sentence and every other input sentence stored in the system in terms of its syntax and its content; ..." (and so on). Note the phrase "syntax and its content"; this refers to the fact that the program has at this stage an internal representation of the correct language syntax as well as the actual contents of the relevant sentences. Consulting the dictionary once more it can be seen that the words "syntax" and "meaning" refer to quite different things; therefore, "meaning" cannot be used as a synonym for "syntax".
Furthermore, nowhere on p. 95 do I declare that the program does not understand meanings. Not only is there no reference to any such statement on this page, there is none in the entire section of which p. 95 is a part. As a matter of fact, what I do say on that particular page is actually quite the opposite: "MCES has the ability to evaluate contexts (in this case sentences which can be linked to each other by virtue of similar words, some more important than others) and combine them into a set which is then used to construct a response to some user input". I then proceed to explain further about such sets and their linkages. I do believe that if criticism should be made then it ought to deal with something that actually exists.
Para. 7 states that "there is very little evidence to previous work in the field". As far as the machine model is concerned, this is correct. And considering the fact that the program stems from an approach which only here and there has been tentatively suggested, it is little wonder. In terms of the human model then the only references to similar considerations were concerned with a rather narrower focus on certain aspects of human thinking, such as perceptions of honour for instance, behaviour in children, or the Müller-Lyer effect, to name only a few. They all have been listed and referenced in the text, with the additional explanations my own model is able to offer.
To see that as detrimental to the overall value of the thesis however is turning the situation upside down. Not only have I shown that the dynamics of the thesis' model coincide with the findings of those researchers, taking the opposite direction it can also be seen that the specific results of the research integrate with the wider domain offered by the model; surely a double confirmation for both parties concerned. All in all, Part One features eighteen references which directly relate to and either implicitly or explicitly support my considerations regarding human thought structures. They are, in terms of their numbers appearing in the text: 1, 2, 4, 5, 6, 9, 10, 11, 12, 13, 14, 16, 17, 18, 19, 20, 22, 23 (pp. 50-52). Given what could be termed the unified field phenomenon above I do not consider eighteen references of this nature as "very little evidence" at all.
As to getting the program to work or not to work (para. 8) - all I can say is that it works on my PC at home in DOS mode under Windows 95, and it also worked under DOS mode in Windows 98 at Griffith University, Technology Building, Room 1.35. My first supervisor, Grigoris Antoniou, ran the program there himself once during the course of the second semester. Obviously this was not the final version of the thesis, but an ongoing development of a program can hardly mean a further deterioration in its performance to the extent that it does not run at all.
Whether the results given in the document seem impressive or otherwise to somebody is not for me to say. I can only point to the fact that the approach taken in the human model has gone further than anything which I could find in the above mentioned references and that there is nothing which disproves my ideas, but plenty of evidence to support them; moreover, that the computer program is a prototypical implementation of the human model, and while not complete shows promise in terms of the relevancies to thought structures outlined elsewhere; further, that the implementation contains features which combine the advantages of both classical and connectionist systems, while minimising the disadvantages of either; and that, from a technical viewpoint, the program exhibits robustness and integrity, as an examination of the code will show (although I doubt whether over 19000 lines of code can be analysed in two days) - the examiner's comments notwithstanding.
I will restrict my concluding remarks to the final section of this letter; the same goes for the second report.
Response to Report 2
In its very first lines the report states that the "student does not know, has totally misunderstood or has chosen to ignore advice given by supervisors as to the purpose of the document". It then goes on to say that while any rules in this regard may not constrain the content, they are about the form of a treatise. And further, that a correct format allows "independent confirmation (or refutation) of the content material".
There are two aspects to that criticism. Firstly, I have read the document on CIT's web site regarding the format of Honours theses. Apart from mentioning the typographical format such as margins, font size, etc (to which I have adhered) it also points out the need for a properly argued, well-referenced presentation - there is no need for me to repeat it all and indeed, the very same line is taken by the report's author when explaining the rules and their reasons in para. 1.
I entirely agree, and I would like to refer back to the first page of this letter, last paragraph, where I summarise these rules, particularly within the context of the present thesis because of the appreciated need to provide confirmation every step of the way and not to allow conjecture and presumption to lead the way. I do take exception to the statement by the author however (still in para. 1), "For the thesis to be valid the argument must be valid and again this is mainly of form rather than content". I do not deny the need for a certain form, but the message here seems to be that provided the form is correct the actual content of the argument is negligible. This is a rather odd proposition and one that syncs with a certain bureaucratic attitude which says that as long as the right forms have been filled out everything else is fine, no matter what the underlying substance may be. The flipside of this approach of course is to say, "I don't care what such and such says, I just don't like the way it has been said."
In para. 2 it is written, "There is no foundation of ideas, concepts or facts that are known or accepted as true to act as a starting point."
The entire Part One is the starting point. As to whether the ideas and concepts can be seen to be acceptable, or true, the reader is presented with corroborating evidence from various fields to show that, as novel as the entire model may be, there is nevertheless research which agrees with the viewpoints and/or conclusions drawn under that model. All the references used in Part One, but particularly those listed under my comments about the first report (see second paragraph, p. 7, this letter) serve as the ongoing confirmations to the line of argument I am presenting. Further, each of those references is not merely placed there as a block of text, but is commented upon to show in what way the reference is relevant and therefore how it supports my own stance. Regarding the relationship between the corroborating evidence and the ideas presented in the thesis, I refer again to the second paragraph on page 7 of this letter.
Para. 3 describes the first part of the thesis as "lengthy, unorganised, and rather self-indulgent". By this stage of the letter I feel I have made enough remarks about the reasons for including something like Part One for my position to be clear on this matter. The Human Model is "lengthy" because of the various strands which I felt pertained to the concept. If it is "unorganised" then I profess to be ignorant about the correct sequence of listing items which exist under the headings of sociology, neuroscience, philosophy, etc. When it comes to the structure within the various sub-sections I don't agree with such a description because I do provide examples, confirmations, and references with any statement made. I am not sure what "self-indulgent" could mean. Is it with reference to the fact that I happen to have a concept which I introduce here; is it my listing of supporting evidence from other sources; or is it for reasons of some perceived superiority on my part for actually daring to introduce a concept and going on to implement it in a computer program? Rather than being "just a disorganised collection of comments about a number of models", the material addresses the various concerns raised by AI research, all with the focus on the dynamics of thought structures. To appreciate this however a reader would have to be acquainted with the relevant literature, where such concerns can be readily found. If I was required to supply the full repertoire of these considerations in the absence of such background knowledge, how "lengthy" would the thesis have to be then?
In para. 4 the author describes the second section as seemingly "irrelevant although lengthy". Once again the criticism confuses length with completeness, and to assert - as I do - that a presentation of an idea should be complete for reasons of objectivity is a notion which ought to be recognisable as valid.
If The Human Model contains intellectual material from a variety of fields, then Part Two, The Machine Model, is rather more constrained. First and foremost, it is about programming, and as such it must be seen as a technical document. To outline a technical process in detail so that every step can be ascertained by a third party is not only desirable; it is also conscientious and downright ethical. I refuse to argue that point any further because to do so would imply dubious attitudes on behalf of the reader of this very letter. Surely this cannot be the case.
The author states, "The program seems to be a re-implementation or perhaps an extension of an existing program called 'SOM' (or perhaps 'KSOM') but the original is not described nor is it clear what changes or extensions may have been made, let alone why."
There is a very good reason why the original is not mentioned: it doesn't exist. What is referred to as "SOM" is in fact an algorithm described in a paper by Kohonen and Hari in connection with investigations into the owl mid-brain as well as patients with category-specific recognition disorders (pp. 23, 24). Why that paper is mentioned can be readily seen by going to page 23, but nowhere is it mentioned in the thesis that the particular algorithm is being extended etc. and used in the MCES program. The relevance consists in the underlying approach, since a type of neural network, called the Kohonen Self-Organising Map (KSOM), has been assumed to be useful for the current purposes.
The "KSOM" the author mentions in the same breath with the "SOM" is mentioned on p. 53 in terms of selecting a particular processing architecture as far as neural networks are concerned, and during the course of this evaluation several variations on the theme are provided in the text (incidentally, this being the reason for having supplied a table of general neural network types in the Appendix, p. 146). In the very next section (2.2., Improvement of the basic KSOM, p. 56), I do discuss the possible shortcomings of the "normal" KSOM and outline certain improvements. One half of these improvements has been instituted as part of a separate project run concurrently with the Honours thesis, and so I merely provided the reference in the text (p. 56) because I did not feel it to be proper to have it featured a second time for an assessment (by the way, to demonstrate the validity of my findings and conclusions in the other project I supplied extensive results from a series of test runs on a KSOM neural network which span over 13000 pages; lengthy - maybe, ethical - definitely!). In any case, a KSOM is not a "program" at all, rather it is a type of artificial neural network which contains certain features and which provides certain results.
If it is not "clear what extensions have been made" to the network type then the cause may be a total unfamiliarity with section 2.2 on p. 56. Not only are the reasons for a further improvement given (added to the modification made during the other project), the changes themselves are described in detail, including the algorithm, and what the improved results are like. To show the exact nature of these improvements the printouts of those results are reproduced in the appendix, starting at p. 147, in the form of pre- and post-modification. To claim, as the author does in para. 6, that the appendix contains "what appears to be some files of data or possible execution results", for which the relationship "to the rest of the document is difficult to see", is quite baffling. To repeat: it is section 2.2, Improvement of the basic KSOM, p. 56, which contains all there is to know about the modifications, the data files, and what they are meant to show.
Further in para. 4 the author writes, "There is a subsection that is apparently meant to be a description that would allow one to use the program" (my italics). Too right it is a description: it says so in the heading, 2.7. Description of the Program (p. 98). And there is more, in line with what a description ought to be; the full contingent of the menus, screen shots, and commands, as well as a complete set of all the error messages, the data type identifiers that have to be known to the user, some technical data about the program itself, and the contents of the diskettes which were supplied with the thesis. I don't think the word "description" is therefore inappropriate or misleading.
Regarding the author's difficulty to get the program to do anything useful I can only refer to what I said already about this matter in the third paragraph on p. 7 of this letter. Could it just be that had the author not dismissed the "apparent description" it might have been possible for him/her to actually use the program?
Regarding the program itself the report says, "Nowhere are we told why this program was written, what the author has used it for or how its construction or use relates to any purpose or goal."
Apart from the fact that the entire thesis is about the implementation of a concept relating to thought structures, there are in fact explicit statements saying just that. They are to be found, in order of appearance, on p. 4, para. 1 of the Abstract; p. 8, para. 3; p. 34, para. 3; section 2.3, starting on p. 60; section 2.4, starting on p. 64; section 2.5, starting on p. 69; section 2.6, starting on p. 94; and finally, in section 3, starting on p. 115. Obviously, if Part Two is dismissed as "largely irrelevant although lengthy" by a reader, then the reason for and purpose of the program would indeed be indiscernible.
The author remarks that the Conclusion, as far as he/she can tell, "boils down to saying that a computer program was written and that more work needs to be done." If a reader had sufficient background knowledge of AI and therefore the issues in relation to symbol-handling vs. connectionist systems, then such a reader would understand the significance of para. 2. These things are far from trivial. The debates about these issues and their associated problems can be found in literature on such relatively diverse subjects as artificial neural networks, cognitive science, logic implementation, learning (human and machine), to name but a few. The references throughout the thesis provide a sample of what articles actually do exist.
In para. 6 the writer refers to "the document called simply 'Appendix.'" Why shouldn't an appendix be called "Appendix"?? I have already devoted considerable space in this letter to the reasons why the profile of Queensland's society has been included; no need to be repetitive.
It is further said (with regards to the Profile) that the writer "could find no reference to this material in the main body of the document, nor is there any analysis with respect to any theory or model of society or social structure." Firstly, towards the end of para. 1 on p. 25 the reference to the Profile can be seen, in black and white. Secondly, the entire section 1.2, Levels of Operation (p. 11 onwards) contains just that analysis the author is apparently looking for. As for the comments regarding the data files I refer to what I have already said in the last paragraph on page 9 of this letter.
To summarise I would like to state the following:
I would like to point out that during the entire duration of my IT courses, either at TAFE between the years 1994 and 1996 and then at Griffith between 1997 and 1999, I have not made any references to the above mentioned history, but solely relied upon my study efforts at those institutions. Due to the criticism received now however I see no reason to withhold such information to show that the ideas drawn upon in Part One, The Human Model, are not idiosyncratic and ill-conceived as they have been made out to be by the assessors.
I strongly object to the criticism made in both reports, not only in terms of the items listed there, but also in terms of the general tone in either, since it implies superficial and negligent behaviour when in fact the features used for such a criticism indicate the very opposite.
While I would readily acknowledge any wrong conclusions I have drawn, any flaws in my algorithms, or any assumptions which I subsequently fail to substantiate, in the absence of any of that I sincerely suggest that my thesis be handed to someone who has access to the appropriate background knowledge in order to make an informed judgment. As it is, one does get the impression that when it comes to Honours theses at least, the university employs a policy of maximal standards, rather than minimal ones.
*) The articles and books were written under the pen name Peter Wenger; at that time my son was in his teens and I wanted to keep my family separate from my professional affairs.