Arthur Dief

  • Student

Activity Feed

On November 13, 2018, Arthur Dief commented on Challenge.Gov – A Model for Government Crowdsourcing :

Thanks for posting!

I’m curious if crowd sourcing government initiatives exposes the U.S. to risk of influence from outside players? It doesn’t feel impossible that a foreign agency could submit an entry under a concealed identity. If that entry is selected, the agency may then manipulate exposures that were unknown to the US.

Further, I wonder if the received responses will be skewed towards those actively engaged in politics and therefore may not be the most accurate representation of the target community. For instance, homelessness is a very real problem, however homeless people generally have limited access to a computer in comparison to an HBS student. If I find a challenge by the U.S. government crowdsourcing how to end homelessness in the U.S., I am able to submit an idea, but I’m also less attuned to the root cause of the problem and viable solutions than the people it impacts. There needs to be an element here of also crowdsourcing the selected solution with the people that the government is trying to serve.

On November 13, 2018, Arthur Dief commented on The Brains Behind Olay Beauty Care Product Recommendations :

Great read!

For me, the use of machine learning to increase sales was unexpected given the market trend showed large players losing market share to small independent brands. I would have expected P&G to acquire a smaller skin care brand or create a sub-line of products branded under a new name with new reference to P&G.

I’m interested to know how the software accommodates for poor lighting, different facial expressions, and appearance with or without makeup. For instance, we are more likely to see crow’s feet when smiling rather than straight faced. Additionally, it would be interesting if Olay had users re-submit selfies over a period of time to compare their skin from before and after the product’s use. With the new photos, Olay could also provide updated recommendations. Finally, I’m interested to know how the frequency of the algorithm recommending a product correlates with the product’s price (i.e. does the algorithm regularly recommend Olay’s most expensive products).

On November 13, 2018, Arthur Dief commented on Vojd: filling the Void in Luxury Design with Additive Manufacturing :

Awesome article! Thanks for posting!

I’m curious if Vojd is considering new materials as it continues to develop its 3D Printing strategy. It would be interesting to see Vojd combine their designs with metal 3D printing, and see if they could achieve delicate designs in gold, for instance, that have previously been impossible to achieve.

Additionally, Ministry of supply is already using a process similar to 3D printing to make their “3D Print Knit” products, spanning from blazers to dresses [1]. Can this technology be advanced to use more designer materials, like cashmere and silk, in order to achieve previously impossible designs or add new interesting shapes and structures?

[1] Halzack, Sarah. “How a Custom Blazer in 90 Minutes Just Might Change the Apparel Business.” The Washington Post, WP Company, 29 May 2017,

On November 13, 2018, Arthur Dief commented on Can Kialo turn online shouting into enlightened debate? :

Thanks for posting!

I’m curious if admins and owners are enough of a blockade to prevent internet trolls and fake news from entering the site. I would imagine that people who enjoy the site often lead discussions on multiple topics that they find interesting. If each post has hundreds of responses, it is unrealistic to expect a single person to comb through and fact check every post.

Even if admins only review reported content, it is possible that they will face difficulty in determining what is the threshold for a credible source. For example, while Fox News and CNN are both national news channels, some people would argue the credibility of articles written on either site. So then do you use an external fact checker and relate the accuracy score of the cited source to the readers? Furthermore, how do we ensure admins are equitable and act without bias to remove content from both sides of the debate that does not meet the quality standards?

At some point the shear scale of policing a post could be overwhelming for even the most engaged admin. In this case, is there any sort of tool that could be developed that automatically relates this information and highlights to users (i.e. a notification that reads accuracy score of XX below the comment) so that admins are not required to be the sole guardians of Kialo from trolls.

Thanks for sharing!

Relativity Space was the first to successfully 3D print a full-scale aerospace quality fuel tank, measuring 7 feet in diameter and 14 feet in height over. The process took 7 days of print time, though longer to develop and test their design. Normally to order and receive a tank of that size and quality would take over twelve months [1]. With this technology already in hand, will Relativity Space’s Stargate become the key to unlocking additive manufacturing for large parts? Should they produce more Stargates, reap the profits, and aid other companies to manufacture bulky items? Or is it in their best interest to keep the technology private?

[1] “Ep 8 Printing Tomorrow/Are We Alone?” HBO NOW®, VICE,

On November 13, 2018, Arthur Dief commented on Love in a Hopeless Place: Machine Learning at OkCupid :

Great read!

OKCupid co-founder, Christian Rudder released a book called “Dataclysm” in which he reveals trends in user data. Two points I found interesting were on race relations and attractiveness. From Rudder’s analysis, he found a bias against black users on dating sites. Measures that OKCupid deems success, including how people rate black users, how often people reply to their messages, how many messages they get are all reduced in comparison to other races. Additionally, women generally are attracted to men that are relatively their same age, whereas men tend to rate 20-year-old women the highest on the site [1].

Though both of these trends measure people’s opinions and actions on the website and do not reflect their behavior in reality, machine learning algorithms are being trained from these actions and delivering potential matches with this input data. Are we biasing our algorithms and perpetuating the trends above?

[1] “Online Dating Stats Reveal A ‘Dataclysm’ Of Telling Trends.” NPR, NPR, 6 Sept. 2014,