A blog posting by a long-ago associate has lingered in a Firefox tab for quite some time. The posting by Dan Weinreb is "Why Did MIT Switch from Scheme to Python?" with further discussion on Hacker News. Don't be fooled by the Hacker News title, because the thread is really about the evolution of a computer science (CS) curriculum.
Thirty years ago this August, I arrived at MIT, and registered in the course 6.031, The Structure and Interpretation of Computer Programs. The lecturer was Robert Fano, an electrical engineer famous for his work in information theory who moved into computer science in the 1960s, at least a decade before the first undergraduate CS curriculum emerged. My recitation instructor was Christos Papadimitriou, who was already distinguishing himself as a theoretical computer scientist. Within weeks, as I listened to the magic delivered by the Italian- and the Greek-tinged voices, I decided to major in computer science.
In the next few years Hal Abelson and Gerald Jay Sussman would concentrate 6.031 into 6.001, the very renumbering punctuating an essential victory. They would pack Algol and Lisp into an elemental Scheme, unifying key concepts and intensifying workload in a curricular tour de force. Yet, my initial reaction was I quite liked seeing the two parts separately for their own individual elegances.
I can remember Fano, with a forceful enthusiasm, explaining call-by-name as he dashed out a diagram with contours and the magic of thunks as chalk hit chalkboard in the Green Building lecture hall. And in the second half of the course the same crystalline clarity emerged in Papadimitriou's recitations of Lisp's dynamic forms and list structures and meta-forms of representation and computation (that's data and control abstractions). As 6.001 came into force, my worry was that the sheer grind and the forging of axes into a volume would lose the potency of the original. However, over the years, hearing wows and even avowals of love for 6.001, I accepted that perhaps not much was lost.
The present shift, with the retiring of 6.001 and the introduction of 6.01 this last year, is much more significant, belied by the simple loss of a zero. The course may still contain a good chunk of what was there, but it hits at perhaps one level of abstraction too high and with an abundance of grit and whirl of "real world" engineering and so much more stuff. All considered, I doubt the MIT CS curriculum will be damaged by this. Still, the transformation brings me to active discussions over the last few years in ACM venues on the declining enrollments in CS programs and on the shaping of computing courses for other fields and the primary education system. You see the same vector of replacing elemental formulations with ones that accommodate teaching students in other fields by contextual-izing with relevant or engaging problem domains.
In all this discussion of pragmatic factors and educational theories, I wonder whether what may be lost in translation (for a while, anyway) is exactly the magic at the heart of the field that draws people into it. Thirty years ago the field's own core curriculum was still being formed for those in the field itself and now with the computing crescendo of the 1990s, we are in the next era of cultural diffusion. Even if it might make sense for MIT to retire 6.001, is it not time for an analogous course for a broader audiencesay, simply Computingto emerge? (Certainly both Peter J. Denning and Jeannette M. Wing have called along this direction for many years.) Just as in high-school courses in calculus and physics, distilled and pure forms may not just be a good idea as a foundation, prior to utility, but also for their innate beauties that inspire and subscribe. And though certainly the voices that I heard are to be thanked for my own arrival, by now it's clear that the forms themselves possess a voice of their own.
Will Rogers once said, "Advertising is the art of convincing people to spend money they don't have for something they don't need." According to Will, advertising is inherently deceptive, and most profitable when it hoodwinks people into paying more for something than they should.
Another view is that consumers lack information and advertising can provide information. In this view, the opportunity for deception only exists because of missing information about reputations and alternatives. If advertisements are relevant enough to inform consumers, then opportunities for deception fade.
Ultimately, whether deception or relevance is more profitable to advertisements may depend on margins versus conversion rates. Deceptive advertising tends to have high profits on each sale, but usually very low conversions. Useful advertisements will yield much tighter margins, but have a much higher volume of conversions.
An extreme example is email spam, horribly deceptive advertisements with awful conversions. And the data there may give us some hope. One of the worst forms of deceptive advertising, email spam appears to be a barely profitable enterprise despite its ubiquity. This may suggest that deception does not inherently maximize profits.
Another example is search advertising. By targeting advertising closely to search keywords and intent, companies like Google have not only made search advertising very lucrative, but also relevant and useful to searchers. In search, ads are highly targeted, rarely deceptive, only occasionally annoying, and often helpful.
But most other forms of advertising remain irrelevant and annoying. The most common technique we see still is broadly blasting ads across all eyeballs. It would be good to make advertising more helpful, relevant, and useful to people. Is it possible?
For a few years now, I have worked on personalized advertising. Personalized advertising tries to make advertising more useful and relevant to people by targeting ads to individual interests and needs.
Recently, I have been struggling with a moral question. Let's say we build more personalization techniques and tools that allow advertisers and publishers to understand people's interests and individually target ads. How will our tools be used? Will they be used to provide better information to people about useful products and services? Or will they be used for deeper and trickier forms of deception?
For me, it is an ethical issue that cuts deep. If personalized advertising will not be used to benefit people, to improve the usefulness of advertising, then I want no part of it. It seems clear that personalization can make ads more relevant, but I fear it also could be possible to use deep knowledge of individual interests to target deceptions. Which will advertisers do? Which will be more profitable?
I am hopeful that we can improve advertising and that advertising will be most profitable for most advertisers when it is useful and relevant. I am hopeful that any deceptions will be marginalized by a flood of more useful alternatives. I remain hopeful that advertising can move toward a helpful information stream and away from the art that Will Rogers deplored.
But, to be quite honest, I sometimes have doubts about the answer, which is why I bring it up for discussion. What do you think? Is advertising an industry fundamentally fueled by deception? Or is advertising better understood as a stream of information that, if well directed, can help people?
I think in the perfect case of a recommender system, advertisements as such become irrelevant. You'd have a direct mapping between customer desire and product awareness. The onus of sales would be on producing products that people wanted rather than on the ability to generate awareness for them.
Naturally, that's not how reality works, but it might be useful, seen as one of the extremes of a continuum. The other end of the spectrum is creating a market for things people don't actually want.
The latter sounds like something that's probably not good and the current state of affairs is somewhere between the two. It would seem that if you can admit the way that things currently work isn't evil, the case of targeted advertisement is increasingly less so.
The problem, of course, is that the function isn't entirely continuous. Targeting can potentially be used to trick people. It's not so much that tricking people is novel in advertising, but as you learn more about how a person ticks, your trickery could be that much more potent.
Patrick Wheeler
Very few ads are intended to deceive. Most mass-media ads are designed to appeal to people's subconscious desiresto be attractive, secure, etc.and to link a product to that desire. Online ads will dominate advertising only if they are effective at making that linkage. Personalized ads ought to be effective at selecting audiences for their subconscious desires by analyzing their conscious choices. See the BBC documentary "The Century of the Self" (available on Google video), especially segment three on the rise of lifestyle marketing, for an example of using conscious answers to infer subconscious desires.
If there is a continuum, it may be from informational ads with conscious appeal, to influential ads with subconscious appeal. Personalized targeting ought to improve the effectiveness of both kinds.
Note that we may still be squeamish about improving "influential" advertising. It doesn't make subconscious desires conscious; rather, it adds a subconscious link to a productand thereby makes us willing to buy more of it and spend more on it than if we had not seen the ad.
Ken Novak
©2009 ACM 0001-0782/09/1100 $10.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2009 ACM, Inc.
No entries found