Artificial intelligence designed business cards are better than yours. Get one designed to show others who you really are.

AutoCard automatically designs personalized business cards based on the things someone likes on Facebook. Facebook likes have been used to compute personality and philosophical traits of the user (e.g. big five)[1,2], which can then be applied to a parameterized model of the business card to provide users with an ideally designed result. Parameter weighting  coefficients were learned by iteratively presenting users with weighted randomized cards and allowing them to select their preferred design. Psychological traits are assumed to transcend preferences for cards, between users.

This project aims to highlight challenges with automating design that stem from issues in design and machine learning. While it’s important to make creative decisions well, we often use poor mechanisms to do so in design. Computational approaches have been applied to this buts its very hard to do so with an appropriate level of depth, so computational decision making often ends up being very naive.

The project is currently not available for public use but I’m working toward making it easily accessible.


  1. Farnadi, Golnoosh, et al. “How are you doing? Emotions and personality in Facebook.” Proceedings of the EMPIRE Workshop of the 22nd International Conference on User Modeling, Adaptation and Personalization (UMAP 2014). 2014.
  2. Youyou, Wu, Michal Kosinski, and David Stillwell. “Computer-based personality judgments are more accurate than those made by humans.” Proceedings of the National Academy of Sciences (2015): 201418680.

Understanding people better than we do

Michal Kosinksi visited CMU today and gave a talk entitled “Predicting Personality from Digital Footprints” in which he describing a mechanism he’s developed for providing (sometimes) better than human ratings of personality over a range of common traits (the Big Five).

I found this pretty interesting as it describes a very effective mechanism by which the notion of philosophical grammars could be formed. Additionally, he and his collaborators have created a public instantiation of their tool with both datasets and a live tool which can be used for testing etc.

Hopefully I will be able to use some of what Michal has done as part of the engine in my project.

The abstract provided for the talk is:

Personality traits form a key driver behind people’s behavior, cognitions, motivations, and emotions; therefore, assessing others’ personality is a basic social skill and a crucial element of successful social interactions. However, based on a sample of over a million participants, I show that personality judgments made by computers―and based on generic and pervasive digital footprints (Facebook Likes)―are more accurate than those made by participants’ friends, family members, and even romantic partners. Furthermore, compared with humans, computers achieve higher inter-judge agreement and superior external validity (i.e. are better at predicting life outcomes). In some cases, computer-based personality judgments are even more valid than self-reported personality scores. I conclude by discussing the consequences of computers outpacing humans in this basic social-cognitive skill.

A critical future for design

Personally I believe that automated design is an important part of our future, but one that is not truly available to us for the time being because two claims I believe to be true: design is not good enough, ML and AI are not good enough. In particular, I feel there is a rift in the design academic and industry communities around the rigor and formality of design, and a similar challenge between in the AI and ML communities in which methodology often trumps meaning.

In this project I hope to respond to these issues by building an artificial (fake) automated design system that embodies many of the critical problems in both relevant fields. I will show the implementation but also make artifacts available for participants to view and provide an interactive context to show how this automated system may replace them, badly, someday soon.

The exhibit I hope to produce will include a laptop or iPad running my system as well as a number of 3D printed failed design examples for a commonly designed object, a chair.

The problem with the singularity is something else

To me, discussions  about the singularity are rather petty, not because its right or wrong, but because it does not matter, and neither the proponents nor the opponents of the concept seem to have useful things to say about that. Elon Musk seems to be actually doing something about it, and, given his track record, I think that’s what’s going to matter.

This is a slightly interesting topic for us to consider because it is somewhat worthless to focus on design fictions that argue about the singularity in the same way it is not that useful for computer scientists to discuss it, instead, our fictions need to consider how we are going to deal with the issues of the impacts of what is happening and what will happen anyway (which we need not call the singularity, but will probably involve computers causing us problems).

In other words, I don’t care if the singularity is completely true, some of it almost certainly is, and we should think about what that may actually mean (though fiction), as opposed to what projections within that field indicate it will mean based on hand wavy science.

Philosophical Grammars for Automated Design


As Ira Glass has pointed out when discussing creativity, understanding and learning to act on one’s sense of style is key to doing successful creative work. This sense of style represents an underlying philosophy which today is a predominant way to convey someones true response to the world. Given tools to interpret personal philosophies with significant efficacy, design determinism could feasibly increase to a point at which automated mechanisms can embody specific designers perfectly and produce artifacts on their behalf. In this world designers would be tasked with an ongoing challenge of accurately representing their changing philosophical responses, and would face competition based on the purity of their philosophies, and the effectiveness of their representations. In other words, designers success would depend on quality and coherence of their philosophical grammars.

Measuring and Representing Personal Philosophy

Journals, public design decisions, social interactions and all other aspects of our living fossil record can lead to near infinitely nuanced representations of our personal philosophies, however, at this time we don’t have a great way to understand this or make sense of it yet. Today search and credit card companies know so much about us that they can sometimes draw more revealing conclusions about our opinions than we can ourselves, however, they tend not to know why and they have almost nothing of interest to do with that information due to shallow representations.

Computational solutions may exist eventually, but for the time being, we can move forward with something more like survey on personal philosophy. Something similar in theory to the MBTI or Pymetrics, except built to review very specific philosophical preferences. For example, with knowledge about if a designer prefers constant radius fillets or curvature continuous fillets may have strong implications to how they will design physical products. On the other hand, how they lean in an empathically motivating situation may offer indicators about how they will design services, systems or experiences. Over time, an approach like this would become more nuanced, eventually encompassing everything we can know about someone to attempt to make as accurate models as possible.

Grammars are often used to provide a codified representation of the structure of language [Chomsky, 1957], but also in design, usually in the form of shape grammars [Gips and Stiny, 1972; Stiny, 1980a, 19980b]. Grammars serve as a potential conduit to convey personal philosophies, but to do so, a formal notion of abstract structure and abstract grammar must be proposed. For the time being, Machine Learning and Algorithmic Game Theory provide a basic formalism for this kind of thing, but a more specific structure will have to evolve.

Implications on Modern Manufacturing

Currently, rapid manufacturing is widely available but tools to support personal design are clumsy and most consumers have little or no understanding of how to ensure a design is good. Assuming we can measure and make use of personal philosophies, rapid manufacturing becomes more flexible as services could offer, for example, to read all your email and provide idealized designs, suited for your philosophical disposition.

Eventually this may mean that designers are supplanted by mind reading robots which create perfect products without much thought, but it will also mean that good designers become philosophical thought leaders as the valuable insights they spend time imparting into their grammar may be beyond the conception of others. In other words, taste and style will remain a valuable commodity, because people don’t only not know what they want, they also, often, don’t know what they think about what they want, before having experienced it.


Philosophical grammars of famous designers and artists would become very valuable to the point that companies would keep them as trade secrets, and industrial espionage involving steeling or somehow interpreting a brand grammar would be a constant consideration between companies like Apple and Samsung. A black market of grammars would emerge in which by espousing ones self to valuable philosophical lessons you may be able to operate at a more mature level and have more carefully designed objects in your environment.


Implementing philosophical grammars for automated design is perfectly plausible with current technology. It is however, relatively repugnant to society and could lead to a pre singularity demise to humanity. Specifically because a system like this is not artificially intelligent, but instead, artificially makes human philosophy as powerful as possible, and for now, this is not something we can handle as a race.


  1. Chomsky, N.: 1957, Syntactic Structures, The Hague: Mouton.
  2. Gips, J. and Stiny, G.: 1972, Shape grammars and the generative specification of painting and sculpture, in C. V. Freiman (ed.), Information Processing 71, North-Holland, Amsterdam, pp. 1460-1465.
  3. Stiny, G.: 1980a, Introduction to shape and shape grammars, Environment and Planning B, 7, 343-351.
  4. Stiny, G.: 1980b, Kindergarten grammars: designing with Froebel’s building gifts, Environment and Planning B, 7, 409-462.

More Than Words

Several years before iPhone and Siri, a friend and I started working on a voice based computer platform. We realized that it could be done locally, or at distance, with computer speech tools or with an efficient human based engine. This approach was as flexible and as powerful as we could imagine any computing system could be, given that humans were the intended users. We went on to collaborate on other things [1] but didn’t have the positioning to make our system work at the time.

Today, this idea seems a little less exciting. It’s a lot more common, there are a lot of systems all around us that do it relatively poorly, and we see that it has some limits, for at least the foreseeable future. However, we may also realize a greater limitation. Even if it were possible to do this perfectly, or at least as well as it’s done in Her [2] and other examples, all we can do is speak. As far as we can tell, for the time being, there’s no faster, richer communication that we can partake in. This is a little depressing, because it means we may be just a few years away from a state of peak communication efficiency with computers, and yet, we probably still need so much more.

Bret Victor [3] has toyed with a version of this issue and tried to consider how the tools we use to achieve things could be designed to help us do more, more deeply, and more rapidly, than we can now. In the same way that the industrial revolution deeply changed society, Bret proposes a notion of a dynamical representation of thought to let us think about systems and their relationships to understand implications of decisions over time, in a way we have a lot of trouble doing now [4].

Sadly, I don’t know of any other examples of this type of view, and in particular, I don’t think I’ve seen it reflected in thoughts about future interactive systems in popular media or in design fiction literature. Similar to the discussion of what do we do when AI becomes more mature than we do, I wonder how we deal with needing to communicate with ourselves and our computers, better than we can with each other.


  1. Shuman, Yosef, and Mark E. Whiting. What Is Service Design? N.p., 2014. Web.
  2. Her. Dir. Spike Jonze. Perf. Joaquin Phoenix, Amy Adams, Scarlett Johansson. Warner Bros. Pictures, 2013. Film.
  3. Victor, Bret. Bret Victor, Beast of Burden. N.p., Web.
  4. Victor, Bret. The Humane Representation of Thought. Vimeo, 2014. Web.

Quotidian Oracle

Designers are notoriously bad at predicting the future through research methods, in no documented cases of major technological or societal change due to innovation, have design methods been the primary motivator in finding a new need or technological opportunity [1]. Although, as he points out, they are very good at making things better. On the other hand, people tend to be great predictors of the future in a wide range of cases. So, while I agree quite strongly with Kinsley, Kirby, and Dourish’s claims and analysis, I feel it’s also quite obvious (and has been for many years) that our media drives our expectation of the future, our technology development, and in many cases career decisions of a large subset of the population. I also feel that while these papers discuss an interesting philosophical phenomenon, there are rich quantitative opportunities for evaluating this and that these would be a great directions for future research, or perhaps research about 20 years ago, when initial longitudinal results of these trends could have been assessed.

Reading these articles brought me to consider how hard it might be to draw conclusions about historic responses to technology or technology related media as our frame of reference is rather skewed by the current one. That said, I feel it would be another interesting exercise to try to better understand how what the public thought effected what technology came into existence at various points in history around the world. Interestingly, current variance in technology on earth suggests this type of analysis could provide really fascinating results. For example, the expectations of wireless phone infrastructure in various African countries was inextricably bound to other forms of infrastructure in the public view, and solutions for a range of related services became available around the same time through technological workarounds, bypassing the slow moving industries in many more technologically robust nations in the West [2, 3]. Mobile banking is one such example.


  1. Norman, Don. The design of future things. Basic books, 2009.
  2. Chipchase, Jan. “Mobile banking: Agents as mediators.” CGAP Blog (2010).
  3. Chipchase, Jan, and Panthea Lee. “Mobile Money: Afghanistan.” innovations 6.2 (2011): 13-33.

An Ideal Forecast

I often wonder how design fiction will really help us. These readings provoked me to wonder, what’s the ideal forecast timeframe relative to industry trends etc.? In other words, I think some forecasting serves to improve our ability to design, while other forecasting may be too focused on being critical or fictional, or perhaps aims to assess a distant future that is too far off to be useful. So, I think working out how to get good value from our forecasting would be an interesting endeavor. Several more particular questions emerge:

  1. How far into the future can we look before we are strictly limited by what we think today?
  2. How much do current trends impact our future design fictions?
  3. What types of design fiction tend to serve in this way best? (e.g. mundane life, catastrophic scenarios, technical fictions, implementations etc.)

I don’t know how to answer these rigorously, but an analysis of past forecasts in comparison to how reality turned out, and, how designers responded to those forecasts, might help us draw some pretty good conclusions.