Senior Thesis Ideas

[liam-thesis-thoughts-blog]

"Final" conception of and reasoning behind Work-Energy Theorem

 Although the route of creating a completely free-rein sandbox environment where students could cause different objects and forces to interact as they see fit appealed to me, i am realizing that going a more instructive route could have a bigger impact. Thus, for my tool regarding the Work-Energy theorem i want to simulate the situation of a ball being raised straight up by a hand, with questions posed that get at the essence of how the Work-Energy theorem is useful for solving problems involving force applied to an object. More specifically, i want to help students understand the constraints involved in analyzing a situation where a force is applied to one object, which then moves a second object. I found that these situations are commonly the subject of the Work-Energy problems on SmartPhysics, the system used by the University for PHYS 211 (and 212).

The focus in the simulation will be a hand holding a ball, initially prostrate on the ground. As some of the arm's energy is used, work done on the hand raises it, lifting the ball as well. The hand stops at an established height, and the ball rises until all the kinetic energy it had gained during ascent is converted to potential energy. Then it falls back down, pushing the hand down slightly when it comes back into contact with it.

Two of the quantitative aspects of the page will be:

Energy transfer bars representing the energy the arm has to give, the KE of the hand, and that of the ball

Background has distance markers so it's clear when arm has risen 1 meter

The energy transfer bars are meant to convey the essential fact of the Work-Energy theorem: the amount of work applied to an object is equal to the change in that object's kinetic energy. Using the metaphor of liquid transfer, i'll convey the idea of the  exchange of energy from the arm to the hand/ball in the form of work. 

The background distance markers will mark discreet points where the KE and work done are to be measured, either by the program or by the student in the form of a question. Using only discrete 1 meter increments allows for more natural numbers in the calculations, serving to scare students less, and allowing them to work the equations mentally so that they stick better. 

One of the biggest additions to my original conception of the simulation will be question-answering being built in. The user will be presented with a question such as "What is the final speed of the hand/ball combo?", and if they do not know the correct formula needed to solve for that quantity, they are presented with a modification of the simulation meant to give them a bit more information as to how the different elements interact. For example, they may be asked to "Quadruple the force the arm gives, and observe the change in the final speed". With enough of these they can reconstruct the formula, in this case W = ∆KE.

[liam-thesis-thoughts-blog]

Current progress on simulation library:

core animation cycle functional using callbacks in proper way to ensure objects always in sync

collision detection between objects works, just stops the objects by default now

Goal for one week from now, before going to break is to have the response to collisions working, force arrows shown for each object, the initial work vs. energy simulation diagrammed. This will allow for mid-April trials on participants

[liam-thesis-thoughts-blog]

The first draft code for drawing a ball with some initial velocity falling due to gravity, as well as that ball's kinetic and potential energy is complete! It was accomplished by extending the drawing library Raphaël to add a fn for creating a ball with a certain location and initial velocity, then hacking together the animation cycle. After some research, i chose Raphaël from among a few popular drawing libraries such as paper.js because it is the most compatible with the devices i want to support and because of its focus on interactivity. Definitely needs refinement, but it's a start.

[liam-thesis-thoughts-blog]

Yian Yian Oh paper about the role of algorithmic explanations and animation in learning

I collected my thoughts on a paper i read last week for summarization here. The papers was "Using visualisations in secondary school physics" by Yian Yian Oh.

It mentioned a study (Monaghan & Clement, 2000) where the students using a simulation about relative motion that used animated feedback did better than the ones getting algorithmic/numeric feedback. It also said that the latter group ended up using faulty algorithms to solve their problems.

So should the relevant formulas be shown in my simulation after all? What would be the effect of showing purely physical or non-symbolic simulations and asking questions that similarly require no arithmetic or algebra? Because learning the equations that relate different properties is so important to traditional physics education i would probably not want to go the purely visual route. However it might be interesting or even more effective to have students start their exploration of the simulation without any formulas visible. Then once they have a feel for the relative importance of different variables like mass and velocity the specific relations would appear, confirming or upsetting their intuition.

Good news from this article was that in Jimoyionnis & Komis, 2001, it was found that using visualizations to teach physics helps students to overcome cognitive constraints and develop a more scientifically appropriate understanding of physics. 

Bad news was that the results from Gilbert, 2005; Gilbert, Reiner & Nokhleh, 2008; Korakakis, Povlatou, Palyvos & Spyrellis, 2009, suggest that having to manipulate of controls in a simulation puts a heavy cognitive load on the learner, which can result in a hindered ability to fully internalize the visualization. This makes my goal of designing an uncluttered interface even more clear.

[liam-thesis-thoughts-blog]

Beginning To Wade Through the Literature

Was pleasantly surprised at the number of academic articles i turned up by simply searching for "interactive teaching physics". I just read through the article "Problem solving - Physics modeling-based interactive engagement" by Funda Ornek. It seemed applicable because in the abstract the author mentioned using a modeling-based approach to teaching physics at a large state university in the US. And upon further research, model-based teaching "seeks to help students create and test appropriate conceptual frameworks which they can reform and add detail as they gain experience and knowledge" according to David Hestenes of Arizona State University, and this is similar to the methodology i wish to employ for my physics-teaching project.

This paper introduced me to some of the terminology of the field of physics instruction and model-based learning that could be useful in guiding my own project. For example:

The right types of questions to use in my own simulations would be their so-called real problems, which require more than one step and the combination of information from different steps in order to be solved. 

The tool would aim to move students toward an expert's style of physics problem-solving, where a qualitative physical explanation is used to guide oneself toward the correct solution, instead of just the application of mathematical formulas.

The fundamental principles in physics models are the linear momentum principle, the energy principle, and the angular momentum principle.

Think-aloud problem solving was used by the researchers to ascertain the level of expertness subjects demonstrated in solving their assigned problem.

They mention p-prims, which were investigated by DiSessa and refer to simple explanatory ideas that go unquestioned and sometimes get applied in the wrong circumstances. Many of the p-prims examined here could be prime candidates for clearing up through my simulations.

[liam-thesis-thoughts-blog]

Mockup Image of Interface

Here is a graphic i threw together to give an idea of what i'm imagining for the visual style of the simulations.

What i would keep for the final style is the asymmetry seen here, like in the ladder rung spacing and grass placement. Obviously it would be easy to generate highly symmetric objects in code, but i'd like to avoid that as i think it could make using the interface rather boring. 

At the top of this image you can see the equations pane, where each variable is expandable and contractable. The simulation parameters could be tweaked to show more or fewer equations. The equation for the maximum angle the ladder can lean before slipping in a given scenario is to be found by the student.

Mockups of what interaction with this simulation might look like to come in a future post.

[liam-thesis-thoughts-blog]

Instantaneous Experimentation

I've had a couple weeks to consider where it is that my tutees need the most help in physics, and the possible origins of their confusion. I also took time today to check out some of the content on the SmartPhysics site, which was created by three physics teachers here at the U of I. One takeaway from my brainstorming has been the idea that a principal goal in creating my initial simulations — and then from there the tools to create new ones — should be lab-like experimentation.

I wouldn't be seeking to replicate the experience students get once a week in their lab sections though. Instead, what i'll be aiming to create is an environment for "instantaneous experimentation": self-directed tweaking and prodding supported by automatic data-keeping that encourages students to discover the laws they're learning about on their own. I think the easiest way to elaborate on how exactly this would be done in practice is with an example.

Students in PHYS 211 here are currently learning about rotational statics; that is, the conditions under which a system involving torque will be stable and unmoving. One illustrative example given to students is that of a ladder leaning against a building. The pre-lecture video students watch on SmartPhysics explains to students that the maximum angle the ladder will be able to make with the wall before slipping along the ground and falling (assuming no friction on the wall) is related solely to the coefficient of static friction between the ground and the latter bottom, related specifically by the equation tan θ = 2µ.

This relationship could probably not be guessed by most students. Instead, they would normally undergo the process of algebraically deriving that relation, using the appropriate equilibrium conditions. Of course they should learn that method because of its extensibility and usefulness in test situations; however, i'm proposing a more exploratory way for them to discover that relationship for the first time. 

I believe we need to allow students to try tweaking all the parameters that are actually tweakable in a given situation and see for themselves the result. Or, just as importantly, the non-result. Perhaps a student's original intuition is that the heavier the ladder is, the more likely it is to slip at a given angle. This is incorrect (somewhat unintuitive, right?), but there is no harm done if the student spends a few seconds tweaking the ladder weight parameter of the simulation to see the non-effect. This tweaking would be done by grabbing, within one of the relevant equations being displayed onscreen, some end of the M variable representing the mass of the ladder and pulling it outwards to expand the variable itself. (See previous post for my thoughts on equation placement.)  Some accompanying animation in the simulated environment would indicate the increase in mass, as well as an increase in the force on the ladder due to gravity, the normal force from the ground supporting the ladder, the force of static friction from the ground, and finally the force of the wall pushing back on the point where the ladder leans against it.

"So", the student hopefully thinks, "increasing the mass just increases each force felt by the ladder by the same factor." Now they're ready to try something else to get that ladder to slip. It is now that they strike upon the idea of decreasing the ground's friction. And so they reach up to the equations, shrinking µ with a simple pinch. There is some animation such as a decrease in the number of blades of grass on the ground that the ladder's standing on, and eventually the ladder slips and falls. Conclusion? Less friction means the ladder slips more easily. Obviously. The next step in the student's hunt for understanding is to find the equation relating the coefficient of friction parameter to the angle at which it slips.

The formula relating these two quantities can be found experimentally in our setup if the student simply gets out a piece of graph paper and records data points with "coefficient of friction" on the x-axis and "angle where the ladder begins slipping" on the y-axis. Except that's not so simple, and boring for most students. So instead we add a piece of graph paper to the virtual scene and allow them to just change the friction parameter. When they choose to take a data point the animation moves the ladder about, finding the maximum angle it can lean it at before it starts to slip for that specific coefficient of friction. The corresponding x, y point is noted on the graph paper. When the student has taken enough data points that they can start to see the relationship, they are then tasked with doing some algebraic work with the applicable equations in order to derive a concrete equation describing the relation they feel they are seeing. 

This process is involved enough to get the student really thinking, trying to dissect a phenomenon they don't yet fully understand in the way a scientist would for the first time. Still, we simplify things for them by letting the animation liven things up and by having software take care of some of the more boring aspects of scientific discovery. This method of "instantaneous experimentation" could be very useful for building intuition about formulas for students. If it is found to help students i could build the library in part around generating scenes where this method of learning is the goal. A professor would sit down in front of the final program and choose "Experimentation" as the type of simulation they want to create, from among other options i shall come up with next brainstorming session.

[liam-thesis-thoughts-blog]

Designing a Physics Simulation Creation Library/Tool

I've decided that the best fit for my senior thesis project would be researching visualization libraries and writing a new library for creating web standards-based physics simulations which could be used by professors teaching Mechanics (PHYS 211) here at UIUC and by others to design simulations to be used as a pedagogical tool for building up a student's physics intuition. 

My reasons for going this route:

physics intuition is not highly valued by students; they prefer to try to memorize which equations apply where

as a tutor with the Center for Academic Resources in Engineering, i am getting a good idea of what are the most misunderstood aspects of physics classes here

most of the physics simulations to be found online are Java applets and have a cluttered visual style

building a scene based on user input that has flawless real-world behavior is a difficult and interesting problem to attack

The deliverable i would aim for by the end of the thesis period would be an extensible library that allows for a user (usually a teacher) to generate, either through an API or by a UI, visually clean interactive simulations run by Javascript within an HTML5 canvas where the laws of mechanics apply as they do in the real world. Physical information about each element in the scene (i.e. position, speed, acceleration, mass, energy) would be easily discoverable through a simple interface. The essential components of this library and the design considerations that went into building it would be documented in an academic paper. I would approach the design only after immersing myself in the academic literature surrounding the design of visualizations and visual languages, which i am already somewhat acquainted with after completing Karrie Karahalios's Social Visualization (CS 467) course.

Outline of path to deliverable: 

(ongoing) studying literature on visual style and visualization languages/libraries

taking notes on what concepts students seem most confused by during my physics tutoring sessions

making a few interactive simulations to find out what ground my library would need to cover

get usability feedback on those examples from students i tutor, document any isolatable change in understanding from using the simulations, with their consent

refine the layout and functionality of simulations the library generates

create a the library and its API for generating the JS and HTML code of new simulations

create an additional UI with potentially more limited functionality for making new simulations for use by teachers without programming knowledge

My main focuses in developing the simulation's (student-facing) UI and look/feel will be:

0 bugginess on interaction of objects etc; the simulation canvas should feel like a window into a virtual world where the real laws of physics apply

encourage exploration and experimentation by allowing for as many modes of interaction and usage styles as possible; simulation doesn't limit user to performing only the phsyics problem meant to be simulated

clean, direct visual style; one potential source of inspiration would be this Röyksopp music video, though my simulations would likely remain 2-D

annotation of information and control over parameters that lets students focus on the concepts behind the equations

That final goal has its roots in two different presentations by Bret Victor: his talk "Inventing on Principle" and a recent blog post of his called "Learnable Programming" (in fact i foresee many of the projects from his site serving as sources of inspiration when designing my simulation UI). I envision the basic UI as having an equations pane off to the left or above the main simulation window. The size of each variable in an equation will denote its magnitude, which will be adjustable by clicking and dragging on the variable to expand or contract it. Any relevant side effects will be indicated on objects or paths within the simulation, as well as in other equations.

For example, the simulation might be showing a cannon ready to fire a projectile. Upon changing the size of the theta (angle) parameter within the v_x = v*cos(theta) equation, the student will see a parabolic path outline on the screen begin to widen or narrow, as well as the cannons tilt adjusting. The theta in the v_y = v*sin(theta) will also be seen to change. These changes could even be applied mid-animation so that a ball careening along one path has its position adjusted so that it is shown where it would be along the adjusted parabolic path at the current time after firing. I will outline more example simulations and features that i have in mind in a future post.

Accomplishing all the above would, i think, mean a great new tool with myriad applications.  Still, i can picture a variety of ways to extend the base project described above which i would attempt given enough time:

add functionality for building Electricity and Magnetism (PHYS 212) simulations

add testing mode for the simulations where a student is asked to predict the behavior of the simulation or choose parameters that will achieve a given aim

optimize the experience for the iPad or other tablets, potentially by building a dedicated application, if there would be advantages over just running the normal HTML5 + JS app in the mobile browser 

I hope to get feedback on my ideas, either in the comments here or through email.

[liam-thesis-thoughts-blog]

Quantifying Human Well-Being

Watching this TED talk by Hans Rosling during high school, i was startled to hear him state that quality of life was actually improving for many living in Africa. If i had been asked before hearing that, i would surely have assumed that life in the 21st century is even harder (and in new futuristic ways) for the poorer populations of Africa than it was in the 20th. However, judging by certain key metrics like infant mortality and lifespan, it appears that their lot is currently improving. 

My pessimistic attitude seems to be shared by many American citizens; a good portion of them actually believe the world will be ending in a few months! The problem likely stems in part from how bad news about violent crimes or corruption always get priority over pieces highlighting how good we all actually have it here in the US. My goal with this thesis project would be to use clear visualizations of cleverly-chosen data points to make a highly quantitative argument about how life in 2012 compares to life at various times in the past. I would try to augment the most traditional and obvious metrics with some unexpected ones that could give a more intuitive understanding of how far our quality of life has come. Obviously, if the metrics i choose end up indicating that life is in every way worse today i will not hide that fact, but i do expect that there are at least a few ways in which lives have improved.

Some initial ideas i've had for what might be useful metrics:

violent crime incidence

percentage of population that can afford to shower every day

quantity of food buyable with an average day's salary

items owned per family (not a sign of progress to all i guess)

different types of food for purchase at average grocery store

distance walked to school/work (problematically low in the US i'd wager)

number of families that lose a member due to non-birth-related diseases each year

cost to see live music performed

percentage of population that visits a beach recreationally each year

Now, organizations or indices that attempt to in some way quantify human happiness on a per-country basis exist already. Some examples are: Gross National Happiness, Satisfaction with Life Index, and the Happy Planet Index. However, these all focus on country-to-country differences and an ordered ranking of all the countries. Looking at their methodology could give me some great ideas, but i would instead try to focus my research on the evolution of happiness from the past to the present. In addition, the result of my research would be a series of visualizations meant to demonstrate the evolution for each metric and not a single index. This means more openness with what went into quantifying happiness/QoL, potentially making my conclusions more convincing to viewers.

The US would be my main focus at first with this thesis, but i'd really like to visualize what sort of evolution has happened in other countries, especially those traditionally seen by Americans as poor. Another possible way to extend the thesis' scope would be to establish some real-time data sources that are indicative in some way of happiness and throw up a website showing the evolution of happiness in a given country up to the present moment.

[liam-thesis-thoughts-blog]

Choice: Interactive educational materials or computer-aided song decomposition

Considering solely the names of the two potential areas of research that i must choose between as they are written in this post title, i immediately feel a draw towards the first. In my opinion, education is the next large field of human activity set to be radically transformed by technology, and such change could potentially bring huge gains in worldwide well-being. I would love to help push the revolution forward in some small way. 

Some projects i've been considering that apply to this field:

Interactive physics simulations written in JavaScript for in-browser display with HTML5. For example, show how the different sizes of the gears make the wheel spin faster or slower with one rotation of the pedals.

Cinemagraphs for iBooks Author or other highly-visual eBooks. Add a little extra prettiness to chapter title images by making them subtly animated in the style of cinemagraph animated GIFs. Most useful would be a way to generate extra frames of some element of the image in different positions in order to have enough frames to make an animation.

Interactive sheet music for music teaching or for when sheet music is used as a graphic in a book like in Gödel, Escher, Bach. 

A whole suite for creating interactive graphics in the style of the ones above...

None of those particularly stand out as "the one", however. One that immediately grabs my interest though is the idea of computer-aided song decomposition. The idea with this would be that in order to create a programmed music video or match the beat in a song with another medium it is useful to have just the drums track or just the vocals track of the song. But how can one go in and extract just a particular instrument's part. Humans are very good at hearing each element separately when they try, and we're also adept hummers or whistlers, so perhaps the computer could extract the bass line from "Money" if i were to hum out that part of the tune to it. It would use the waveform i generated to guide its search within the song. I believe it would be very possible to at least isolate the rhythm of a particular track of a song using a method like that. 

Well, i'll have to give it more thought. The draw with education is strong, but i also really want to have the right tool to make highly-programmed music videos where visual elements move with distinguishable tracks.