Radu Jianu (radu.jianu@city.ac.uk), project supervision

I am a lecturer in the giCentre at City. My research interests are in Data Visualization, Visual Analytics, and Human Computer Interaction. I try to understand how people interpret and extract meaning from visual content (generally data) and what makes such visual content 'good'. I'm also interested in how people interact with digital technology and in where new technologies can improve established workflows. I run user studies (quantitative, qualitative, enthnographic, observation), use eye-tracking technology, mine data using visual methods, design visualizations, and create public facing websites that show data in engaging ways.

I'm in in A401C (access from the 3rd floor of the College building on a metal staircase; knock on the door at the top).


Past supervised projects

I typically supervise students doing Data Science and HCI though I have also supervised a few students on GamesTech, Sofware Eng. and IT courses. Here are the titles of projects I've supervised in the past two years The average mark for these projects was approximately 67%:
2017/18

2018/19

When interested, students are encouraged to publish their results as academic papers. For example, here is one dissertation outcome that will feature as a short paper in the CHI2019 conference (the premiere HCI conference): CHI2019 Commercial nudging.

Broad categories of projects

(types of projects that I tend to supervise; specific topics are listed further down)

Observation studies (most suitable for students in HCI and IT) Pick a technology (e.g., chatbots, fitness apps, banking applications, ...) that you find intriguing or are a fan of and do a qualitative observation study: get people (e.g. 15-20 of them) to use one or more instances of the technology and collect data about it (e.g., through in-the-lab observation, ethnographic observation, or diary studies). Analyze the data using formal methods (e.g., thematic analysis, distributed cognition). Focus on general 'what works and what not' or on particular technological aspects (e.g. persuasive technology, collaboration, social interactions, trust).

Alternatively you can focus your observation study on a real-life context for which no adequate digital support exists yet (e.g., how could video learning resources be augmented to better support collaborative learning); or on a particular technological or social aspect (e.g., persuasive technologies; belief changing; trust; information foraging).

Interested? Scroll down and take a look at some topics I suggest; even better, pick a technology or context that you are excited about (e.g., Do you like to exercise and use fitness apps? Do you like to travel?) and then chat with me to find a suitable research angle and methodology).

Quantitative (comparative) user studies (most suitable for students in HCI, DataScience+Vis, GamesTech): Pick two alternative technologies (or ways of) doing the same thing, have people use them and measure how effective they are.

Depending on your interest and abilities you may consider evaluating:

For such studies it's often the case that you need to implement yourself the stimuli (i.e. interfaces, systems, visualizations, games) used in your study (because you need full control over their presentation). Thus, to pursue such projects you need to have the skills to create the required stimuli.

Interested? Scroll down and take a look at some topics I suggest; even better, pick a visualization or an aspect of interfaces/games that you are excited about and then chat with me to find a suitable research angle and methodology).

Interpreting data using visual methods (most suitable for students in Data Science, HCI, or IT) Pick a dataset or construct one yourself, explore it from different angles preferably using visual methods, and report your results. Depending on your skills you can use GUI visualizations tools such as Tableau, or programmatically and you can consider creative public-facing websites showing your results (in a static or interactive way). Some programming skills might be necessary at least to manipulate raw data.

Interested? Scroll down and take a look at some topics I suggest ; even better, think of a dataset that you are excited about or a topic that you think you may be able to get data about. Come and chat with me to find a suitable research angle and methodology).

Implementation or system building (Anyone with strong programming experience) Implement a piece of software that can do something that couldn't be done before.

Interested? Scroll down and take a look at some topics I suggest; even better, think of a your own dream system or app. Come and chat with me to find a suitable research angle and methodology.


Specific topics

(they can be approached with some or all of the methodologies above)

How do domain experts and data scientists communicate and collaborate? Imagine a domain expert (e.g., health professional, economist) working together with a data scientist trying to explore data and figure out answers to some domain-specific questions. The domain expert knows nothing about data-mining while the data scientist knows nothing about the domain. How do they communicate, find common language, refer and point to data elements and results, optimize their use of computing resources, divide their work? This subject could be the target of an observation study.

How do people learn to program (or other types of learning such as languages); a system 1 vs. system 2 and situated cognition perspective: When teaching programming, lectures most often emphasize problem solving and higher level cognition (system 2 cognition). They describe abstract high level concepts first, then expect students to instantiate them in concrete problems. However, what if we consider programming to be a skill - such as tennis or dancing? In this case, learning by doing or from example might be more effective. Is programming a system 1 or system 2 activity? Are there learning tools that support this type of learning and how would we design one that would? This project aims to explore such topics but is fairly open ended. This can be the focus of a quantitative study, a qualitative study , or a system design/implementation.

Developing analytic creativity: what is the space of possible questions a dataset can answer? When given a brand new dataset (and tools to analyze it) people typically run out of ideas of what to do with it quite quickly. Consider IMDB movie data (i.e. full information about movies and casts): a few questions might spring to mind and you might explore them, but how soon before run out of ideas? Would you consider looking into how the prevalance of movies genres evolves over time, the correlations between genres and rating, or the evolution of female actresses over time? Such examples can spark new analysis directions and lead to a better use of data. Thinking about the space of questions that a dataset can answer in a formal way and how it can be used to drive analyses can be the basis of a few different types of projects (theoretical and/or computational modelling, questionnaire/quantitative/qualitative study).

The effectiveness and impact of persuasive technologies Websites and applications (e.g., for learning, for fitness, commercial ones) often use persuasive elements that aim to influence their users' behavior (e.g., encourage participation, buying, etc.). We can use observation or controlled studies to study the impact of such persuasive elements in real applications using ethnographic/questionnaire/diary methodologies.

How do people do visual data analysis? Characterizing the process by using HCI frameworks (Distributed Cognition, Situated Cognition, etc.) The process of making sense of and deciding with data by using visual analysis tools (e.g., Tableau) is complex: users start with high level goals or are guided by features in the data or the visualization; they refine their objectives and test hypotheses; they engage in low level interactions; and use the visual tools to support their cognition. Describing it by using frameworks established by the HCI community (e.g., DC, DiCoT, Situated or External Cognition) is relatively unexplored and leaves space for many interesting projects. These would likely involve conducting observational study with participants using data visualizations in practice and analyzing data using HCI frameworks.

A data analysis of a research area It is relatively easy to get access to all titles/abstracts/keywords/author-affiliations of a research area or topic (e.g., all papers published in the top visualization conference; all papers talking about gender equality). Such data could be mined and visualized to reveal interesting trends (e.g., What are major themes and how did they evolve over time? What are major players in the fields and how are they geographically distributed? Who collaborates with who?).

Data analysis of topics as revealed by social media Similar to the project above but with data collected from public social media streams.

Crowdsourced/collaborative eye tracking annotation: Eye-tracking can tell us where on a computer screen a person is looking when viewing a video. Any eye-tracking analysis will involve mapping 2D gaze coordinates (where on the screen a viewer is looking) to what was shown in the video (the objects, people, etc.). To be able to do this efficiently we need to know the positions and shapes of things shown in the video at each moment in time. Going through the video and drawing shapes around objects one frame at a time is very laborious. It would be great if we could 'crowdsource' this effort - have many people do a little bit of work and combine their results. This project involves building software.

Comparative evaluation of visualizations of eye-tracking data Eye-tracking tells us where people look on a computer screen. Participants in eye-tracking studies switch their focus multiple times a second. As such, eye-tracking data is often large and difficult to make sense of. There are several popular ways of visualizing such data (AOI time lines or scarf plots, gaze heatmaps) but it's unclear which are most effective.

Visualizing IMDB (movie) data (or football dat? or any other data?) IMDB data can be downloaded and visualized to reveal interesting connections between actors, movies, directors, genres. This project is about analyzing the data to find interesting patterns and/or creating a public facing, online such visualization.