9:30 am

Ben Feeser Salience and the formation of memories Humans are presented with incalculable amounts of sensory input in everyday life: visual, touch, sound, smell, and taste. How are we able to pull out the salient information from these vast inputs to form memories? If we are to understand the formation of memories, then we need to understand what makes information salient. Research proposal.

I will perform a literature review on salience and contribute my own hypothesis for what makes information salient.
I will design a test with human subjects to better determine what makes information salient.
I will outline further areas of research to better understand salience and the formation of memories.

Hypothesized what makes information salient, after reviewing latest literature on subject.
Proposed a test that would demonstrate what makes information more or less salient.

Annika Heuser The Best Way to Learn a Language If we wish to acquire another language in the most efficient way possible, we must devise a strategy based on the research conducted by linguists and computer scientists who have created language-understanding models. Research Proposal

I will learn about relevant theories concerning the representations systems that we may use for language. In order to do this, I will read Fillmore's “The Case for Case,” Borchardt's “Casual Reconstruction,” chapter 9 of Jackendoff's book Semantics and Cognition, called “Semantics of Spatial Expression,” and various chapters of the book The Atoms of Language.
I will interview Professor Berwick about computational approaches of language understanding, and I will specifically ask about his research on computational models of language acquisition.
I will interview Professor Flynn, a linguistics professor who conducts research specifically on second language acquisition, about the strategies that she recommends for learning a language.
I will learn about the strategies employed by popular language learning programs, such as Rosetta Stone, Babble, Duolingo, and learning in a classroom setting, so that I may recommend one over the others.
I will propose a strategy for learning a second language based on what I have learned from the previous steps and I will propose a way of testing this strategy.

Compiled the ideas of multiple linguists and computer scientists who have built language-understanding models into a comprehensive overview of what we know about human and computational language acquisition.
Proposed an efficient way of learning another language based on the ideas about which I have learned. This strategy may be employed by readers interested in approaching the task of learning a language in a different way or it may be developed into a new program for learning a specific language, such as Spanish or German.
Recommended next steps for testing how effective the strategy that I proposed is and recommended currently available language learning programs that I think will be most useful in learning a new language based on what I learned from the research that I did.

Nava Haghighi Self-Interfaces: A Window into the Human Body and Brain Imagine what it would be like to know why you react to events the way you do, or what you base your decisions on. We have evolved to mostly notice low-level perceptual inputs and are often unaware of the higher-level processes that take place in our bodies and brains. We often attribute our decisions to tangible external stimulus, but the external stimulus are processed and turned into higher-level concepts once received. What if there was a way to peek into the brain and experience the higher-level translations of the low-level perceptual information? Can a live window into our brains change how we perceive the external world or the decisions we make? Proposal for a pilot experiment.

To test the proposed hypothesis, one must consider the ways in which low-level perceptual information is translated in human brain. Specifically, two categories of interest are physiological signals such as heart rate variability, electrodermal activity, pupil dilation, and respiration, as well as chemicals produced in the body and brain such as adrenaline, dopamine, and cortisol.
1. Select a desired behavior
2. Select a chemical or physiological signal you suspect might correlate with the behavior
3. Provide live biofeedback on the changes in the chemical production or physiological signal
4. Measure behavior change

Proposed the steps necessary to test whether providing window into higher-level cognitive processes can create behavior change.
Identified examples of physiological signals or chemicals in human body and brain and described how they might correlate with certain behavior.
Discussed the positive and negative implications of this theory on what humans currently define as "self".
Described how this work relates to the larger discourse on Theory of Mind such as Minsky's K-Lines theory.

Kritkorn Karntikoon The Effect of Visualization on Problem-Solving Skills If we are to understand the role of visualization in problem-solving skills, we need to understand how visualization helps our problem-solving performances and the limits of this effect. Pilot experiment.

Subjects will be asked to solve some problems that normally require figures such as geometry and physics problems. Subjects will be divided into 4 groups and given the same set of problems in different formats including (1) purely written (2) purely drawn (3) misleading figures (4) figures with too much information.
To guarantee that each subject has enough skills to solve the problems correctly, I will give a set of easier problems that use the same methods and provide solutions before the actual experiment.
I will observe the process or methods that each subject use to solve the problems. I will draw conclusions from this data and propose a conjecture about the effect of visualization on problem-solving skills.
Further experiments might be conducted to verify the conjecture based on the results from experiment. A possible experiment is to verify if equations give the same effect or can be counted as visualization.

Identified new visual routines that humans use to solve geometry/physics problems.
Demonstrated that visualization improve humans' problem-solving performances.
Identified the limits of the effects of visualization on problem-solving performances by determining types of visualization that does not improve the performances.
Distinguished between visualization and abstract thought.
Proposed a theory to describe this phenomenon.

Luke Luneau Improve Genesis for Business Case Studies If Genesis is to be a proof of concept for story understanding, it should be able to understand a wide variety of stories, including business case studies. Pilot experiment.

Decide which type of business case study I want the Genesis system to understand.
Depending on the type of story, decide what Genesis should conclude and want connections should be made.
Depending on the story, decide what rulesets should fire. Add to the rulesets if need be.
Compare the Genesis externalization graph to what I think it should look like.
Verify that Genesis is making a proper understanding of the story.

Adapt Genesis to understand case studies, further increasing the variety of stories the system can understand.
Demonstrate that Genesis can be adapted for business case studies and is applicable in a field not yet explored by the system.

Molly Mason AI as a Collaborator in Architectural Design Meetings and Design Space Problem: Sharing knowledge is hard because each party involved has to recreate the idea in their own mind and hope that it matches how everyone else recreated it. This is especially challenging in the design of architecture because these ideas must be shared through multiple mediums and between multiple disciplines who possess different vocabulary and visualization styles. If we are to use AI to assist design meetings then we must identify the most common mistakes within design collaboration, analyze the transferability of a similar concept between multiple mediums and fields, and strategize procedures of clarification. Such an AI would be able to assist with clarification, coordination, cooperation, and collaboration during design meetings. If this information is collected in the right structure it would have the possibility of being better integrated into existing CAD and BIM tools. Research Proposal

In this project I will: Identify common mistakes and miscommunications within design collaboration, identify different commonly used representations in design meetings, propose a means of mediating between these mediums and describe what qualities a successful AI Collaborator might need.

Demonstrated that collaboration is both a linguistic and medium problem. Analyzed existing design process for structural errors. Recommended steps for how knowledge sharing can be facilitated with AI. Proposed a structure for how a meeting information could be organized. Described how future scenarios could be aided by such an AI.

Jessica Quaye Augmenting Genesis' story understanding using image processing If we are to fully understand stories, then we need to leverage visual information to complement analysis of the written script. Research project with implementation.

1. Run Genesis on the script of two speeches: i) Oprah Winfrey Harvard Commencement Speech 2013 ii) Steve Jobs Stanford Commencement 2005 address.
1. Ask Genesis questions about how the speakers are feeling at different time marks.
1. Take images from the different timed scenes and run them in a CNN to predict speaker's emotion.
1. Analyse consistencies and inconsistencies between Genesis and CNN responses.
1. Since Genesis seeks to model human understanding and interpretation, I will ask 10 people to predict the emotions in the images.
1. If the images do a better job predicting the human interpretations, it might be worth integrating image understanding into Genesis to augment the story understanding process.

Collected and analyzed data to establish whether image understanding can augment Genesis' story understanding capabilities.

Links to speeches Jobs: https://www.youtube.com/watch?v=UF8uR6Z6KLc Winfrey: https://www.youtube.com/watch?v=GMWFieBGR7c Jim Peraino Recipe following and question asking in architectural design If we want to design buildings that make us happier and healthier, then we need a design process that combines human imagination with computational simulation. Hand Simulation

I will propose a simple design task that cannot currently be done by a computer without a human imagination, and outline a simple blocks world hand simulation that demonstrates how a computer could ask a human for help during this process using recipe following.
I will propose a simple design task that cannot currently be done by a human without computational simulation, and outline a simple blocks world hand simulation that demonstrates how a human could ask for help during this process using recipe following.

Framed the problem of incorporating both creative and empirical input during the architectural design process.
Illustrated how recipe following can enable humans to ask questions of computers and enable computers to ask questions of humans during the design process.

Sameena Shaffeeullah Counting Tobacco Hornworm Caterpillars In order to create a robotic assembly line that dynamically puts food in trays, we must know how much food each tray needs by counting the number of caterpillars in each tray. Pilot experiment.

The first step is research ways to count the number of caterpillars in each tray without using any sort of neural network, as there are not very many images of this type of caterpillar available.
The second step is to create a basic implementation of 2 methods of counting caterpillars. The two methods I foresee myself pursuing are (1) thresholding the image and (2) using color detection or color filtering methods (like flood filling or color segmentation).
I will then decide which of these methods look most promising and pursue that one further. I suspect that I pursuing an algorithm that has elements from both methods.

Created an algorithm that can count the number of tobacco hornworm caterpillars in an image without performing expensive computations or taking time via a neural network.
Demonstrated that object counting can be done without deep networks.

Suchan Vivatsethachai Understanding Internal Representation and its Conversion to the External World If internal representations exist, there must also exist mechanisms that convert between the internal representations and the external representations. Therefore, if we are to truly understand the internal representations, we have to also understand the conversion mechanisms and their properties. Pilot Experiment

I will design an experiment with human subjects that will help me gain more insights about the internal represenations and the external representations of humans. I aim to perform this experiment on about 10 to 20 subjects.
Specifically, the experiment will involve the visual system and the proprioception of humans. I will ask the subjects to focus on a target for a certain period of time. Afterwards, I will ask the subjects to close their eyes and touch the target with their fingers. Because the subjects shut off their visual systems, they no longer receive input from the external visual representations. Any behaviors from that point on will result from only the relationships between the internal representation of the visual system, the internal representation of the proprioception, and the external representation of the proprioception.
The target could be a lid of a water bottle for example, and the subjects will be asked to touched the lid with a finger (or two.) If the subjects can touch the lid with high accuracy, then that means there is a mechanism that can translate the visual focus that becomes an internal representation after the eyes are closed into external proprioception. Then I will perform follow-up experiments to strengthen and confirm the hypotheses I make after conducting the first experiment.
Lastly, I will record the result, analyze the data, and make statistic conclusions. With the analyses I make, I will then infer about the properties of the internal representations and the conversion mechanisms between the internal representations and the external representations.

I framed a problem that potentially help researchers gain more insights about the internal representations of humans.
I presented a type of experiment that can be used to investigate the relationship between the internal representations and the external representations.
I tested the speculated properties of the conversion mechanisms between the internal and the external representations.

Rose E. Wang, rewang@mit.edu How Humans Use Their Story Understanding for Multi-Player Game Coordination If we want to understand human intelligence, we must study how humans are capable of quickly coordinating with one another. Previous AI milestones include DeepMind's successful AlphaGo project which demonstrated a single AI system can learn by playing from experience. However, an important milestone yet to be reached is enabling multiple systems to effectively cooperate together, such as for self-driving vehicles or household robots. With this project I hope to investigate how humans use their story understanding skills for dynamically coordinating together across different multiplayer game levels. This project pursues this topic in the context of Overcooked, a cooperative multiplayer game where players play as chefs and must prepare meals to specific orders under a time limit. Pilot study and research proposal

Motivate the question of story understanding: Why should we understand stories?
Motivate the use of multi-player game: Why do we use games as a baseline environment for understanding intelligence?
Identify what story understanding means when playing multi-player games.
Develop experiments which would display our use of story understanding in order to coordinate in a multiplayer game. This would include controlling for game difficulty level, team setup (purely collaborative, or mixed cooperative-competitive settings), game layout, etc. Analyze the types of communication, coordination tactics (sequential ordering of tasks?), formation of distributing responsibilities, etc.

Define what coordination means in the context of multiplayer game playing, and how that connects with real-world coordination.
Motivating why story understanding is important for executing game coordination.
Explaining and demonstrating how story understanding arises in real experiments, which vary in difficulty and setup.

Erica Yuen Representations in Color Perception If we were to understand how we recall the perception of color, we need to understand if multimodal associations have an effect on our ability to remember and differentiate color. Pilot Study

Subjects will be presented with a color on a computer screen. Subjects in the control group will be given 1 minute to stare at the color. Then they will have a 10 minute break. They will then attempt to recreate color using HSV (Hue, Saturation, Value) sliders.
Subjects will be given another color and 1 minute to stare at the color. However, they will also be presented with a word next to the color. The word will also be read out loud by a computer generated voice. Then they will have a 10 minute break. They will then attempt to recreate color using HSV sliders, but with the word presented to them in the same format as they are trying to recreate the color.
The above experiment will be repeated 10 times, for a total of 20 color matchings per subject. I will draw conclusions from this data about whether the subjects that are presented with a vocabulary and representation of the color if they are able to recall the perception of color better.
I will calculate the vector differences between the original color and the recalled color. I will draw conclusions from this data about whether the multimodal presentation of language will improve color perception.

I will and analyze data relevant to establishing whether the presence of written and oral vocabulary aids with the recall of color perception.
I will demonstrate whether the presence of language visually and verbally helps form stronger internal representations of color, and if these representations aid in color perception memory.

11am

Sabrina Drammis Theories of Neural Computation (A Review) If we wish to understand how the brain works, we must start by understanding how information in the brain gets transmitted. Paper review.

I need to refresh and improve my understanding of the cell and molecular biology of neurons. I will read select chapters from "Principles of Neural Science" by Kandel et al.
My first topic of study will be spiking based models. I will read "Models for Neural Spike Computation" by Staelin and "Networks of Spiking Neurons" by Mass.
Next, I will study the topic of synfire chains. I will read Abeles' paper "Synfire chain in a balanced network" and his Scholarpedia entry on synfire chains. I will read the paper "Analysis of synfire chains" by Herrmann et al.
Finally, I will read Bartlett Mel's work on the role of dendrites in neural function computation in "Pyramidal Neuron as Two-Layer Neural Network" and "Diversity and Dynamics of Dendritic Signaling".

Analyzed seminal and leading work in theories of neural computation.
Provided a review paper of neural computation focused on computation at the level of an individual neuron.

Alex Cabral Understanding Self-Models across Different Demographics If we are to understand human intelligence, then we must understand how people model themselves in their own minds. I plan to conduct an online study to learn about people's self models and how those models differ between different demographic groups. Pilot study

I will construct a series of questions that address many of the hypotheses and open issues discussed in Marvin Minsky's The Emotion Machine. Those questions will include ones such as “What are some words you would use to describe yourself?”, “Given that list of words, separate them into any number of groups and give each grouping a name.”, “What physical location embodies the essence of yourself?” “Do you think you are the same person you were 10 minutes ago? 10 weeks ago? 10 years ago?”
I will convert those questions into interactive forms that help get the correct answer.
I will combine the questions into an online study for my research group's platform, Lab in the Wild.
Finally I will run the study with a small group of users in order to identify potential issues and start collecting data.
Any additional time will be used to revise the online study and run it with a larger group of people.

Designed a study that gets people to reflect and accurately report on their self models.
Gathered information about different self models.
Began to build an understanding of self-models and how they will have to be constructed for machines to truly mimic human intelligence.

Nichole Clarke The Effect of a Reader's Bias on Story Modulation If we are to understand the role of bias in modulating stories, then we need to understand how our own biases influence our ability to discern the bias in a story. Pilot experiment.

Subjects will be asked to read two pairs of short articles. Subjects in the control group will first read the pair of stories and answer questions related to the author's bias. Thereafter, they will read the second set of short articles and answer questions related to their perceived bias. The experimental groups will read and answer questions in the reverse order. Each article in a pair will have different biases.
We will draw conclusions from the data (their answers to the questions) about whether the subject's bias affected their ability to discern the author's bias.

Collected and analyzed data relevant to establishing whether subject's bias affected their ability to discern the author's bias.
Demonstrated that the reader's bias impairs (or does not impair) their ability to discern the author's bias and intent.

Erick Friis Leadership and Collaboration in the Genesis Story Understanding System If we are ever to build a collaborative artificial intelligence based on story understanding, we must first understand how models of leadership, hierarchy, and cooperation can be stored as stories and used to influence interactions. Research Project with Implementation

Research existing methods for building collaborative artificially-intelligent agents (even if not story-understanding related)
Get the Genesis problem solver working on my computer and learn how to simulate and instruct multiple agents
Formulate stories that teach functional collaboration principles to agents
Formulate stories that teach dysfunctional collaboration principles to agents
Evaluate how groups of 2-3 agents work together when instructed as such
Present my findings in a 15 minute oral presentation

Identified key problems for artificially-intelligent agents to functionally collaborate with each other and humans
Developed short stories that Genesis agents can learn functional or dysfunctional collaboration and leadership principles from
Proved that different kinds of collaboration principles can be taught to Genesis agents with stories

Henrik Giske Fosse Language's Role in Virtual Spatial Cognition My vision is to understand human cognition, and to do so I need to better understand spatial cognition. A part of understanding spatial cognition is figuring out which parts of the brain are active in the process. Hermer-Vazquez, Spelke and Katsnelson proposed in 1999 that language plays an important part in spatial cognition and I will therefore investigate this further. However, Fedorenko and Varley have disputed the results from Spelke's experiments. I share their skepticism, but because the results are of such importance in understanding human spatial cognition, I will reproduce their experiment in a virtual environment. Reproduce experiment.

I will create a octagonal room with the Unity3D engine similar to the one used in Spelke's experiment. By looking at a corner and pressing the button, you will be able to select a corner. This application will be transferred to an Android phone equipped with virtual reality googles.
10 Participants will wear these virtual reality googles and be shown an image of one of the four corners in the room and are tasked to orient in the room to find the correct corner, and they will do this a total of 20 times. While doing this, the participants will either be shadowing verbally or rhythmically.
I will draw conclusions from the success rate of participants finding the correct corner while shadowing and how this success rate compares to the control trial.

Depending on the outcome of my experiment, I will either have strengthened or weakened the credibility of Spelke's results.
I will have created a program in Unity3D that allows testing of spatial cognition in various environments with simple modifications. It could easily be modified reproduce the experiment where they deconstructed the room into segments as well.

Abhinandan Jain The Effects of Thermal feedback on Aesthetic Chills and Cognition Aestheic chills or Frissons are universal markers of openness to experience in the Five Factor(OCEAN) Model of personality traits. We aim to use thermal feedback to elicit Aesthetic chills and investigate the downstream effects in cogntion towards a stimulus. Pilot experiment.

Subjects will be asked to watch Charlie Chaplin's speech from the movie Dictator. The test group will receive a thermal feedback on their back at specific points in the speech and the control group wont. We will collect physiological data like Heart Rate, ElectroDermal Activity and Facial Heatmap during the experiment.
Subjects will be asked to fill survey asking questions about the chill eliciting points in the speech and their perception of the speaker.
We will analyze the survey and physiological data to determine frequency of chills, the effects of chills on subject's valence/arousal and subject’s cognitive load.

The data will establish the effects of thermal feedback in eliciting aesthetic chills and its downstream effects in cogntion towards a stimulus .

Ronit Langer An Ethical Framework for a Human Intelligence Model If we are to create a model of human intelligence, then we must equip the model with an ethical framework for moral reasoning. Humans communicate about ethics using moral language. Drawing on tools from philosophy and computer science, this project proposes a method to give machines a way to interface with moral language. Assuming computers are given ethical capabilities, this project explores how and to what extent we would want computers involved in deciding our moral dilemmas. A research proposal

Analyze the papers presented in 6.xxx through the lens of ethics, paying close attention to different representations that can be used to encode moral language.
Research and read papers in philosophy and neuroscience about the role of reason, rationalization and justification in decision making.
Propose how reason, rationalization and justification can be encoded in a moral language for a machine to use.
Assuming a moral language, discuss the opportunities and limitations of including machines in our ethical deliberations.

Synthesize ideas from philosophy and computer science to propose a mechanism for encoding moral language into a machine.
Explore the opportunities and limitations of having computers with moral reasoning.

Jennifer Madiedo Predicting Story Outcomes Through Story Alignment using Genesis If we want to understand how we predict outcomes of an event, then we need to understand how we generalialize prior knowledge from past events. Research Proposal

Analyze related work on story alignment and story generalization, and write-up review. Produce a preliminary hypothesis on how story alignment can lead to accurate outcome prediction.
Compile a set of potential approaches to achieving hypothesis with pros and cons of each. Propose reccomended approach with rough outline of steps and expected contributions.
Time permitting, begin basic implementation

Completed review of story alignment and story generalization related work applicable to outcome prediction.
Produced hypothesis on story outcome predication via story alignment technique
Recommended potential approach and rough outline of what steps should be taken to achieve hypothesis

Ole Morten Olsen The role of creativity in problem solving If we are to understand the human mind, we need to understand how we solve challenging new problems by a combination of acquired knowledge and creativity. What is this creativity and can it be replicated in a machine? A research proposal.

Review existing literature (cognitive science, neuroscience and computer science).
Based on the reviewed literature, find or make a definition for human and computational creativity.
Evaluate how machines can behave creative by themselves or be part of creative processes in cooperation with humans.
Make a proposal for an experimental implementation based on the evalation.

Reviewed and analyzed literature about existing ideas use of creativity in problem solving.
A definition of computational creativity.
Proposal for an implementation of an experiment.

Ellen Shea, Temitope “Tosin” Olabinjo Can the Human Language System Multi-Task? Because language plays a crucial role in human intelligence, if we are to understand human intelligence, then we must understand the limits of the human language processing system. Pilot experiment.

Subjects will be asked to complete a total of six truncated SAT sections: three math sections and three reading sections. For the control, the subjects will take one math and one reading with no noise distractions. For one of the tests, the subjects (who understand native-level English) will take the other two of the sections while listening to a podcast in English. For the final test, the subjects (who do not know Hindi) will take the last two sections while listening to a podcast in Hindi.
We will draw conclusions by comparing the results of each trial for each individual.

Tested whether the language system is involved when completing various tasks, and whether the system is activated when one is listening to speech in a language they are unfamiliar with.
Demonstrated whether human language systems can multi-task.
Collected and analyzed quantitative data that describes the effects that background conversations in a known and unknown language have on cognitive tasks.

Tristan Thrush Implementing the Intuitive Motion Planner If we are to understand how humans plan motions in the real world, then we need to understand how we are able to account for and predict the effects of newtonian physics. Project proposal.

Detailed architecture ideas will be provided for a robot motion planning system that has access to a physics simulator (such as Drake: www.drake.mit.edu). The motion planner will be able to use the physics engine and perform probabilistic inference to approximate real world dynamics, and accurately develop humanlike plans for behaviors that an agent without an intuitive theory of physics would not be able to accomplish (such as throwing objects to a certain location). It is important to mention that I also did a project about this idea in Tenenbaum's class, but that project was about introducing the idea and inspiration from human cognition; the motion planner itself was not implemented and architecture plans were not discussed in detail.
Necessary perceptual systems for mapping the state of the world into an internal physics engine to do the motion planning will also be explored.

Designed a system that models the hypothesis where humans make use of an internal physics engine in motion planning.
Will likely use this proposal as a launching point for a project that goes beyond the scope of this class.

Sunny Tian The Effect of Directed Questions on Image Memory If we are to understand human intelligence in the context of forming memories about images, then we need to understand how noticing specific aspects of a complex image increases our ability to recall the image contents later. Pilot experiment.

Subjects will be asked to analyze a complex image. Subjects in a control group will not be required to answer any questions. One test group will be asked to summarize what they see in the image. A second group will be asked pointed, directed questions about particular details in the image.
A week later, we will call back the subjects and ask them questions about the image. Questions will vary in terms of difficulty (detail) in order to determine how much and what aspects each subject actually remembers.
We will draw conclusions from this data about whether the subjects that summarized or were asked pointed details about the image better remember complex details within the image. Due to the nature of the experiment, subjects are provided paper and are allowed to redraw the image to help them visualize and clarify their explanations.

Collected and analyzed data relevant to establishing whether summarization or drawing attention to details aids humans in their recall of images.
Demonstrated that exercising the human language to form mental models of images (through summarization or directed questions) helps focus human attention on details, thereby increasing (or not increasing) a subject's ability to recall images.

Justin Xiang and Nikhil Bhatia The Effect of Auditory Jamming on Reading Comprehension If we want to understand the role of external and internal language in performing cognitive tasks, then we need to understand how different types of auditory jamming affect our ability to utilize our inner language faculty while performing reading comprehension tasks. Pilot experiment.

All subjects (MIT students) will be asked to take a reading comprehension section of the SAT. We assume that all MIT students have college-level reading comprehension abilities and thus the reading comprehension section of the SAT is a fair comprehension task.
Subjects are then broken up into four groups: Control, English, Secondary Language, and Foreign Language. Subjects in the control group are tasked with taking the SAT with no auditory jamming. Subjects will be given 20 minutes to take a reading comprehension section of the SAT (with the vocabulary questions omitted). Scores will be calculated as a percentage of questions answered correctly. All subjects in the control group will be tested on the same reading comprehension section of the same SAT test.
Subjects in the English group are tasked with taking the SAT while listening to a Bible audiobook in English. Subjects will be given 20 minutes to take a reading comprehension section of the SAT (with the vocabulary questions omitted). While taking the test, subjects will be exposed to auditory jamming by a fixed 20-minute section of the Bible audiobook, played at a fixed volume via airpods.
Bilingual subjects in the Secondary Language group are tasked with taking the SAT while listening to the same Bible audiobook in a non-English language that they can understand. All subjects within this group will take the same SAT reading comprehension section (but a different section than the control group and the English group). Based on the secondary language spoken by the subject, we will play the corresponding fixed 20-minute of the Bible audiobook via airpods.
Finally, subjects in the foreign language group are tasked with taking a different SAT reading comprehension section while listening to an audiobook in a language that they are not able to understand at all. We chose Arabic as our incomprehensible language because the Defense Language Institute in the US identifies Arabic as a member of the language group that is most difficult for native English speakers to learn. Furthermore, none of our subjects had the ability to comprehend Arabic. By picking Arabic, we hoped to understand whether an incomprehensible language still triggered human inner language mechanisms the way comprehensible languages do.
We will draw conclusions from these experiments about whether auditory jamming affects human inner language and thus, reading comprehension ability. Furthermore, if it is the case that auditory jamming affects reading comprehension ability, we will be able to draw conclusions about whether the specific language of the auditory jamming plays a significant role in hindering subjects' performance.

Collected and analyzed data relevant to establishing whether auditory jamming affects reading comprehension abilities.
Potentially demonstrate that distracting the inner language faculty decreases (or does not affect) a subject's reading comprehension abilities.
Potentially demonstrate that the level of audio comprehension (comprehensible vs incomprehensible) affects a subject's reading comprehension abilities.

Jessica Zhu Training a Machine to Play Cards In order to understand how machines can work with humans in solving a problem, a first step is to teach a machine to play a team-based card game based on a set list of rules and a varying list of strategies. Research Proposal.

I will list the rules of a specific card game, called Tractor. These rules cannot be broken, and must be followed regardless of who is playing.
I will list some strategies for Tractor, and describe situations when certain strategies are more optimal. Some of these strategies will involve taking in real-time inputs and modifying existing strategies based on those inputs.
I will describe a framework for implementing a machine that is able to play Tractor well. This framework will include pseudocode for outlining the rules and strategies that the machine will follow.

Compiled list of rules and strategies to teach a machine how to play the card game Tractor.
Created framework for implementing a Tractor-playing AI.