Invitation this week’s Graduation Presentations
Posted: 22 April 2019 02:27 PM
Administrator
Rank
Total Posts:  6
Joined  2018-12-11

You’re all invited to this week’s Public Graduation Presentations by Esmée Stouten (Thursday) and Lotte van der Kamp (Friday)

Date and time: Thursday April 25th, 16:00 - 16:40
Location:  Gravensteen Building (Pieterskerkhof 6, Leiden), room 0.11
Student: Esmée Stouten
Thesis advisors: Maarten Lamers, Maaike Harbers and Peter Paul Tonen (Capgemini)

Title: “Exploring How Developers Could Include the European Commission’s Ethics Guidelines to Strive Toward Trustworthy AI”

Abstract: Artificial intelligence (AI) is neither a revolution nor hype. As the potential benefits of AI are increasing, the usage of AI-enabled technologies also engenders certain threats and risks that require a careful approach. Based on a first draft of ethics guidelines concerning trustworthy AI from the High-Level Expert Group on AI, this explorative study attempts to provide critical insight into the way companies could offer their developers guidance regarding the concrete operationalization of trustworthy AI. To this end, during a time frame of nine months, two qualitative studies were conducted. First, semi-structured interviews were held with 16 stakeholders to map current problems and possible solutions within the AI fields of different companies. Subsequently, developers for the IT-consulting corporation Capgemini were asked to utilize the assessment list designed by the European Commission to gain an understanding of how their current working method corresponds with the key components of the original list. The results demonstrate that the subjectivity in defining ethical concepts and the unique view that every person possesses on an ethical dilemma make it difficult to compose a predetermined set of rules for making informed choices that adhere to certain principles. Also, the fuzzy front- end of innovation processes as well as the dynamics within AI-enabled technologies ensure that an assessment cannot be truly comprehensive. Additionally, composing rules when a project’s direction is yet unknown ensures that there can be no stability in this regulation. Besides these content-related difficulties, there appeared to be tension between gaining freedom and taking responsibility; developers usually do not look beyond optimizing technical possibilities. In response to these overarching difficulties, recommendations to company-specific guidelines have been provided; among others, avoid ambiguity by giving meaning to subjective terms, utilize examples to illustrate relevance when considering a specific aspect, and add a weighted-scale checklist to gain insight into the extent to which a system is compliant with ethics guidelines. As little research has been conducted into drafting ethical guidelines regarding AI for companies, it cannot be claimed that different forms of ethical testing are not at least as effective as composing ethical guidelines. It is therefore important that more research is conducted toward applying a system of ethical governance in AI and robotics.


Date and time: Friday April 26th, 10:00 - 10:40
Location: Gravensteen Building (Pieterskerkhof 6, Leiden), room 0.11
Student: Lotte van der Kamp
Thesis advisors: Tessa Verhoef and Marcello Gómez Maureira

Title: “First Person vs. Third Person perspective in a digital Memory Palace”

Abstract: The ancient method of loci is a mnemonic device which allows users to remember items by mentally placing them along a route in a familiar environment. While originally purely real life environments were used, improvements in digital environments in video games prompt the question if the option of viewing an environment from a third person perspective might improve the user’s ability to memorise characters in a digital environment, as opposed to the first person perspective. This research tested two groups. Both groups were presented with a video of a walk-through of a video game environment. Along the route through the environment, they saw 3D models of fictional characters at several locations. The first group saw the walk-through in the first person perspective, while the second group saw it in the third person perspective. Participants were then asked to recall which characters they saw where, first on a schematic map of the environment, and then based on images of each location. While there was no statistical difference between the first person and third person groups, there is a significant result between the schematic map test and the location image test. Participants performed significantly better locating the characters on the schematic map. This could suggest that the Method of Loci works so well based not on the anchoring of memories to other information (such as the item of furniture that the character stands next to) but rather, the actual location in space.

Profile