*** Next offered: Spring'14 ***
Version: Tue Jan 14 19:31:23 EST 2014
Prereqs | Textbook | Goals | Weekly | Projects | Grade
- Prof. David C. Brown
- Course Content
- 2014 Schedule
- References (not all may be used)
- Spring 2014 Class list
- Before the first class please read:
- the web-based paper Intelligent Computer-Aided Design
- The role of AI in Engineering applications is steadily increasing. Systems are being used for diagnosis and trouble shooting, design, intelligent control, planning, and a wide variety of other expert applications. AI will play an increasing and particularly important role in the integration of information technology into Engineering systems. The NSF indicates that there are many research opportunities in Engineering Design.
This course will examine some of the AI-based work being done currently, and in the past, on design problem-solving. The domains will be Electrical Engineering design, Mechanical Engineering design, Civil Engineering design, and Software design (i.e., automatic programming).
The main goal of the course is obtain a deeper understanding of what design is, and how AI methods and AI-oriented thinking might be used to support it and study it.
This course will be of particular interest to those people wanting to prepare for research in design problem-solving or those building design applications. It will also be of interest to those studying CAD, CAE, CAM, and other computational methods of support for Engineering, as well as those interested in AI applications.
Graduate students from departments other than Computer Science are welcome provided they have some CS background. Past classes have included graduate students from Mechanical Engineering, Fire Protection Engineering, Electrical & Computer Engineering, Manufacturing, Civil Engineering, and Computer Science, as well as many people from Industry.
The course will be run in seminar style with readings of current literature and with student presentations. There will be several papers to read each week. There are writing assignments, but no programming assignments.
Prior knowledge of Artificial Intelligence would be a distinct advantage. AI course preparation is strongly suggested, and should only be waived after consulting with the instructor.Those without an AI background should study an introductory AI text, such as Winston or Russell & Norvig, to learn about Search and Knowledge Representations. Help outside class and guidance regarding what to read will be made available.
- Recommended but not Required:
- CS534 - Artificial Intelligence.
- Note that taking this course at the same time as the Grad Intro to AI course is acceptable.
- There is no textbook. We will be using the excellent introductory survey articles from the AI in Design special issues:
as well as many other selected readings.
- IEEE Expert/Intelligent Systems & Their Applications, Vol. 12, No.2, March-April 1997
- IEEE Expert/Intelligent Systems & Their Applications, Vol. 12, No.3, May-June 1997
(Note that these are available in PDF via any WPI computer at: March-April 1997, and May-June 1997.)
Of course, I have to suggest (it is not required!) looking at the following for background reading:
- To survey some of the literature, and to answer the following questions:
- What are the issues in the application of AI to Design?
- How do we define "design"?
- How do we characterize:
- types of design?
- types of design systems?
- How do we critically read a paper about a design system?
- What are the strengths and limitations of approaches to design?
- Why is it part of AI and why is it worth studying?
- How do AI researchers approach such problems?
- Who are some of the "key players" in the field?
- Which are the influential systems?
- How can these systems be compared?
- What are some of the current research areas?
- During the course each member of the class will make at least one presentation, by themselves or with another person (depending on class size). Each presentation will be about a design system or theory, and based on readings from the literature. I will provide you with the relevant paper(s). If you can find more that's great -- just make a copy for me.
You should expect a Presentation to last (at least) 30 minutes. I will try to provide the papers for the research that you are to present, 2 weeks in advance, so that everyone gets the same amount of time to prepare. The earliest talks may only get a week to prepare: but they are simpler systems so that should be OK. You should use Powerpoint (or similar) and be able to download the slides before the start of class.
Each week there will be an assignment of papers in order to prepare you for the following week's presentations. My goal is to put all the papers in pdf on the web for you. You must read all the papers each week. You will all be expected to answer questions about them in class. The reason "why" each paper has been included has been provided online in the Schedule.
At the start of each class (starting with the 2nd) I will collect from everyone a short critical review (Critique) of each paper you've read about the system(s) to be discussed on that day. Each critique should be no more than 1 page. Your critiques will be evaluated and returned in the following class.
You do not need to provide critiques of the IEEE Expert Survey articles, nor of the pre-first-class readings. But you should be prepared to discuss them and ask clarifying questions.
You will be expected to actively participate in class by asking and answering questions.
- Profile Creation Task:
- The ongoing task for the course is to write a "profile" that consists of a set of logically grouped questions that if asked of a design system (or theory) would reveal enough information about it to allow it to be compared with all other (similar) AI-based design systems (or theories).
Here is an example of such a profile (pdf) for Expert Systems in general (i.e., not explicitly about design).
From the questions asked, their organization, and from the sample answers you should make clear what your view is of how to characterize design systems, and what the different types of domain-independent and domain-dependent design activity might be.
You must develop your profile as a web page throughout the semester. Send me an email containing the URL early in the course. Gradually add questions to the profile and refine it each week. Every week you must indicate on that web page the questions that were recently added (or refined) in your profile. Please do this by Friday after each class. I will look at them and try to provide regular feedback.
- Ongoing Creativity Project:
- The National Science Foundation has identified Design Creativity as an important topic for research. This project is concerned with design creativity and will allow you to try to use an AI approach to studying it. The project will have several stages (with the deadlines shown at the schedule page) that should guide you to an interesting result. It is important to be honest about doing each stage in turn, as the journey is as (or more) important as the destination.
The nine stages are as follows. Each stage involves some report to me: even if it is only a short note. In most cases I'm not expecting more than about 3-5 pages, sometimes one, and certainly less than 10.
- Study, evaluate, compare and write up a few key existing definitions of creativity. Make sure that you cite them.
- Write up your initial thoughts about whether design programs can be creative, and how: i.e., is computational design creativity possible? Boden's video might be useful as a stimulous for your thinking:
http://www.vega.org.uk/video/programme/81 -- (I hope this link still works!)
- Produce a list of criteria that you believe people use to evaluate whether an engineered product is creative. i.e., what aspects do they consider when they decide?
- Pick a product that you feel most people would identify as creative and use your criteria to evaluate it. Tell me what you selected and why, then report on your evaluation. Better to keep the type of product not-too-large and fairly common: e.g., not the Space Shuttle.
- For each of your criteria separately, describe how an AI-based design program might produce designs that met that criterion, or scored highly for that dimension. Focus on what knowledge the program would need to have and what reasoning it would need to be able to do. This is hard, so please put some effort into this step!
- Read a design creativity paper that will be assigned to you. Provide a short critique of the paper.
- Paper: Developing computational design creativity systems
- Assessment scheme (for step 7):
- Use a product creativity assessment scheme (to be provided to you) on the product you selected. Provide a short note about that experience, plus the results obtained, and what you think of those results.
- Assess your work on steps 3-5 based on the reading and the scheme used. Describe what was good and bad about your initial reasoning and decisions: i.e., critique yourself!
- Reflect on steps 1-8. Write your conclusions about the possibility of an AI-based design program producing designs that most people would consider creative.
Note that an AI approach would include building a program to test your hypotheses about how a design program might produce high scoring designs, and then evaluating its performance. However, due to the diverse backgrounds of the people that normally take this class no programming assignments have been included.
- Final Project:
- The final project for the course is to use the profile that you've developed during the course to help you describe, categorize and evaluate/critique a journal article that you select. When you select a journal article please let me know which it is and why you have selected it. This can be done at any time during the semester. Earlier is better. The particular choice can make quite a difference to how well your profile works! It is probably better to select an article that reports on a system that does some sort of designing, as opposed to a theory paper.
The article that you pick must come from Vol.25 (2011), Vol.26 (2012) or Vol.27 (2013) of the AIEDAM journal. These papers can be accessed as PDF files online via any WPI computer, as WPI has a subscription. See:
The journal is also available in hard copy in the WPI Library. Note that I may be familiar with some of the articles as I was the journal's Editor in Chief from 2001-2011.
At the end of the course, submit on paper:
- A copy of the journal article;
- The result of your profile applied to the journal article ( < 10 pp approx.).
Make sure that you provide each profile question followed by the matching answer: i.e., don't just provide the answers.
- The grade will be based on:
The presentation(s) will be evaluated by technical accuracy, presentation style/quality, and degree of insight.
- (10%) class participation
- (20%) the presentation(s)
- (30%) the weekly critiques
- (20%) the ongoing creativity project
- (20%) profile creation & final project
- Additional Information:
- 1995 NSF Strategic Planning Workshop for Engineering Design, Position Paper
- NSF Report on Research Opportunities in Engineering Design, April 1996.
- NSF Report - ED2030: Strategic Plan for Engineering Design, March 2004 (pdf).
- Books about AI in Design
- DCB's Design-related Definitions
- WPI Artificial Intelligence in Design Group
- AI in Design Webliography
- "Let's Examine the Definitions of the Terms We Use in Engineering Design Research"
- AI EDAM journal
- George Forman 360 Grill -- I'm a Grill
- Dr Susan Besemer
dcb at cs wpi edu