#12 Semantic distance

Reilly et al. (2023) study how the relationships between neighboring words can help describe the flow of ideas through a text. Their method measures the semantic distance between consecutive content words using two representations of word meaning, one based on human ratings of experiential features and the other on patterns of word use in written English. In a simulated word-list task, both measures detected transitions between categories such as animals and musical instruments. In Jack London’s To Build a Fire, words paired in their original order were more closely related than randomly paired words, especially with the measure based on written English. The study gives researchers a way to examine where a text stays with related ideas and where it shifts as the narrative unfolds.

This article gives me a way to explore how players move from describing what happened in my magic game to trying to explain it. A player might write, “I thought the game only registered clicks. Then it showed the symbol I had been looking at.” The relationship between those sentences tells us which expectation the reveal challenged. If the player adds, “Maybe it used the camera, but I couldn’t tell what it was measuring,” they also identify a specific gap in their explanation. Semantic distance could help me explore where accounts shift from describing the reveal to proposing a cause. I would pair it with close reading, because two sentences can use similar words while expressing very different degrees of certainty. This connects to my central research question about how LLMs recognize and use people’s uncertainty when interpreting their explanations.

The feedback from this week’s check-in and workshop was to keep my current storyline and start bringing it to life visually. I took the main advice to be that I should get a complete version working before trying to perfect every details. That gives me a way to move forward with the story and see how the scenes work together.

I’m meeting with my committee chair this week and plan to confirm the next deliverable then. My immediate production milestone is a first animated version of a short sequence from the game. I’ve designed the 2D character and am ready to start animating this week. On the research side, I’ve narrowed the language question to how players express what they expected, what they observed, and what they still can’t explain. I’d like to discuss how to capture those distinctions in the retelling and follow-up questions.

Reference

Reilly, J., Finley, A. M., Litovsky, C. P., & Kenett, Y. N. (2023). Bigram semantic distance as an index of continuous semantic flow in natural language: Theory, tools, and applications. Journal of Experimental Psychology: General, 152(9), 2578–2590. https://doi.org/10.1037/xge0001389

#11 The Construction of Impossibility

In “The Construction of Impossibility: A Logic-Based Analysis of Conjuring Tricks,” Smith, Dignum, and Sonenberg (2016) examine how magic creates the experience of witnessing something impossible. They argue that this experience depends on the whole sequence of a performance, rather than a single moment of distraction. Their account distinguishes the sequence spectators believe they have witnessed from the actual sequence that makes the trick work. Apparent impossibility emerges when the final outcome conflicts with expectations formed from earlier events, yet spectators cannot resolve that conflict. The authors illustrate their framework through a logical analysis of a close-up trick. This is a theoretical account of how magic is constructed, rather than an experiment measuring audience responses.

This distinction helps me connect the creative and empirical directions of my project. In my narrative game, an agent predicts a participant’s choice through concealed eye-tracking. I need to consider what participants believe the system can observe at each stage, because a correct prediction alone may not create a magical experience. For the artwork, the gap between the visible encounter and its hidden operation invites reflection on our dependence on technologies we only partly understand. For the study, it helps distinguish participants’ observations from their explanations. I will collect both, alongside their confidence, perceived impossibility, and recognized knowledge gaps. LLMs will then interpret these accounts with or without the additional self-reports. The article supports my decision to treat perceived impossibility separately from confidence: someone might find the outcome difficult to explain while remaining confident in a particular explanation. It also gives me a way to review whether the narrative prepares the reveal clearly.

This week, I finalized the current empirical study design in preparation for meeting with my committee members on Thursday. The meeting has not happened yet, so I do not have feedback to report. I would like to discuss whether the participant questions clearly distinguish observation, explanation, and awareness of missing knowledge, and whether the planned comparisons can answer my research questions.

For my proof of concept, I am developing the narrative arc, the guide character, and the overall visual presentation. I plan to create a short playable sequence that takes a participant through the setup, private choice, and prediction reveal. I already have an eye-tracking prediction demo; the next step is to give that mechanism a coherent experience around it. This proof of concept will help me assess the pacing, clarity of instructions, and relationship between the character’s behavior and the reveal before expanding the full website.

Reference

Smith, W., Dignum, F., & Sonenberg, L. (2016). The construction of impossibility: A logic-based analysis of conjuring tricks. Frontiers in Psychology, 7, Article 748. https://doi.org/10.3389/fpsyg.2016.00748

#10 Magic Before the Reveal

In the prologue to How We Became Posthuman, Hayles (1999) examines how the Turing test frames our understanding of human and machine intelligence. When people judge an unseen participant through responses on a screen, bodily presence becomes separated from its digital representation. Hayles questions the assumption that intelligence can be understood independently of embodiment. She also compares the test to a magic trick: assumptions accepted at the beginning shape how people interpret what happens afterward. The important transformation occurs before someone decides whether they are communicating with a human or a machine. By participating, they have already entered a system that connects their perception and actions with a technological interface. This shifts attention from the machine’s apparent intelligence to the conditions under which that intelligence becomes recognizable.

This perspective helps me reconsider where the magic begins in my project. The final prediction is the obvious trick, but the earlier instructions, drawing activity, and journey through the website also shape what participants expect and notice. Eye-tracking provides a technical mechanism, while the narrative gives participants another way to interpret the encounter. I want to consider how the interface makes some parts of this process visible and leaves others unclear. For example, a participant might understand that the camera is active without connecting their gaze to the agent’s prediction. This creates a gap between participating in a system and understanding how it works. For the media art component, that gap matters beyond whether the trick succeeds. It offers a way to reflect on everyday technological dependence through an experience of uncertainty. In developing the project, I want to ask what participants believe the system can do before the reveal, and how the reveal changes that understanding.

Finished in-class activity: Project description

My project uses an interactive web-based magic game to explore how our thoughts and beliefs are shaped by systems and shared biases. We witness a seemingly impossible effect without understanding how it is produced, just as we rely on technologies we do not fully understand. Through this experience, I explore the shock, fear, and uncertainty that shape our encounters with technology and our criticism of it. Inspired by the human condition and the limits of what we can know, the project asks how we live with insecurity when complete certainty may remain out of reach.

Project progress

I did not make progress on the project this week. I am still thinking about how to connect the magic experience with the broader question of living with technologies we only partly understand. Next week, I plan to review the website’s opening sequence and clarify what participants are told, what they do, and what they might infer before the prediction. I also want to refine the questions asked afterward so participants can describe their experience separately from their explanation of the trick.

Reference

Hayles, N. K. (1999). How we became posthuman: Virtual bodies in cybernetics, literature, and informatics. University of Chicago Press. https://press.uchicago.edu/Misc/Chicago/321460.html

#09 When agreement does not mean accuracy

In “Extracting the Truth From Conflicting Eyewitness Reports,” Waubert de Puiseau et al. (2012) examine how researchers can reconstruct an event from several imperfect accounts. Thirty participants watched one of two versions of a video that provided different amounts of information about an argument. They then answered questions about the event. The researchers compared majority voting with a model based on cultural consensus theory, which estimates differences in witnesses’ knowledge from their response patterns. The model reconstructed the event more accurately than majority voting in the tested conditions. However, its assumptions become problematic when people give similar wrong answers because of shared misinformation or expectations. This distinction matters: agreement can reflect knowledge, but it can also reflect a common mistake.

This article helps me clarify the research question behind Trust and Know. In my web experience, participants will encounter a magic reveal and explain how they think the guide predicted their choice. Several participants might offer the same explanation because the performance directs their attention toward similar clues. I want to examine whether AI can recover the actual mechanism from these accounts, especially when their interpretations are misleading. To make this test useful, I plan to ask participants separately what they noticed and how they explain it. Someone could misunderstand the mechanism while still reporting an observation that helps identify it. I also want to distinguish trust in the guide from confidence in an explanation and actual understanding. Giving the AI the same accounts with and without these ratings would help test whether the ratings improve its inference or distract it from useful evidence. The study therefore gives me a starting point, while my project extends the question to open-ended explanations of a deliberately misleading experience.

The feedback I received on my preproduction binder was basicly positive, with no specific revisions suggested. Preparing the binder helped me organize the experience into a clear sequence and connect the storyboard to my research questions. My next step is to turn the storyboard into a working demo and check whether the drawing, calibration, and magic reveal fit together smoothly. I also need to refine the questions that distinguish participants’ observations from their explanations, and their trust in the guide from their understanding of its mechanism. I will discuss these choices with my committee as I prepare for the pilot.

Reference

Waubert de Puiseau, B., Aßfalg, A., Erdfelder, E., & Bernstein, D. M. (2012). Extracting the truth from conflicting eyewitness reports: A formal modeling approach. Journal of Experimental Psychology: Applied, 18(4), 390–403. https://doi.org/10.1037/a0029801

#08 Clarifying What Trust Means in My Project

In “Trust in Automation: Designing for Appropriate Reliance,” Lee and See (2004) review research on how people develop trust in automated systems and how that trust influences their use of technology. They explain that trust becomes especially relevant when a system is too complex for someone to understand completely. A central distinction is between trust as an attitude and reliance as behavior: feeling confident in a system and choosing to follow its recommendations are related, but they are not identical. The authors also emphasize that greater trust is not always desirable. Trust should correspond to the system’s actual capabilities; otherwise, people may depend on it too much or reject useful assistance. Their framework considers how system characteristics, context, and cognitive and emotional processes shape these relationships.

This article helps me clarify a problem I am facing in my project: what exactly would count as evidence of trust after the magic interaction? My demo uses webcam-based eye-tracking to predict a participant’s choice. A successful prediction might make the system seem impressive, but that does not necessarily mean the participant would rely on it in another situation. I am considering whether a later decision task could help me examine this distinction. However, if that task involves a gamble, participants’ decisions could also reflect their attitudes toward risk or their understanding of the game. I would therefore need to examine their reported trust alongside their choices. The article also gives me a more specific question about machine authority: could the ritual and magic presentation encourage participants to attribute abilities to the system beyond what the demonstration actually shows? This connects my creative idea to a research question without assuming that the goal is simply to increase trust.

This week, I continued working toward a focused research question and a workable experiment, and I updated the demo. I have completed all the training modules last time. I didn’t meet with committee members this week. My main challenge remains connecting the creative interaction with a clear experimental comparison. I’ll do this next few weeks and meet with my committee then.

Reference

Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human Factors, 46(1), 50–80. https://doi.org/10.1518/hfes.46.1.50_30392

#7 Midsemester Reflection

  1. My research progress has not followed exactly the direction I imagined at the beginning of the semester. I started with a broad interest in machines, magic, and ritual, but translating those ideas into a specific psychological question has been more challenging than I expected. I think the project will continue to change as I refine the research question and experimental design. One milestone I feel good about is developing a magic demo that uses webcam-based eye-tracking to predict a participant’s choice. This gives me a concrete interaction to build on. I have also developed an initial experimental idea for studying trust in human-computer interaction, although I am still working out what kind of trust I want to measure.
  2. I would say my project vision is about 70% of the way toward what I hoped to establish by this point. The technical development has gone reasonably well, but the visual design remains open and will likely change as I clarify the interface and the participant’s experience. I do not yet feel that I have a finished production milestone I am especially proud of. The demo is encouraging, but I am still struggling with how the proposed trust-based gambling task connects to the magic interaction and the larger research question.
  3. For the accompanying paper, I want to explain how theories of media ritual and machine authority inform my design choices, then examine whether the experience works in the ways those theories led me to expect. The creative format offers a way for participants to reflect on human-computer relationships through their own experience. The paper can connect that experience to the literature and analyze participant responses, including findings that challenge my initial assumptions.
  4. In the second half of the semester, I am most looking forward to establishing a focused research question and a workable experiment. I want to make the connection between the creative experience and the trust task clearer, so that each part serves a purpose. I also hope to finish a testable version early enough to pilot it, make revisions, collect data, and leave enough time for analysis.

#6 From Everyday Media Rituals to an Encounter with AI

James Carey’s ritual view of communication left me great impression when I first learned in undergraduate class. This week, I revisited chapters from his book Communication as Culture, including “A Cultural Approach to Communication,” which connect with my project’s idea. Carey explains how communication creates and maintains shared meanings. He distinguishes this ritual perspective from a transmission view, which focuses on sending information across distances. Using newspaper reading as an example, he shows how an everyday media practice invites people to participate in a particular representation of the world. Readers encounter stories that organize social life into recognizable roles, conflicts, and values, helping sustain a shared understanding of society. In this account, ritual extends beyond religious ceremonies to ordinary practices through which people express and reaffirm shared beliefs. Carey also explains that the two perspectives can coexist: the same activity can provide information while maintaining a shared understanding of reality.

This chapter helps me clarify what I mean by ritual in my MA project. My interactive web-based experience would combine an AI agent, guided breathing, and an interactive magic performance, potentially incorporating eye-tracking or other physiological signals. I want to explore how participating in this sequence might shape the relationship a visitor develops with the system, the AI agent and the authority they attribute to it. Carey’s perspective encourages me to consider how the whole encounter establishes meaning: the opening question, the instructions, the visitor’s actions, and the final reveal could all contribute to how a computational output becomes believable or personally significant. I am especially interested in whether a temporary sense of trust can emerge while the system’s mechanisms remain unclear. However, an everyday media practice develops through repeated participation, whereas my project may involve only one brief encounter. This difference inspires me to investigate: which aspects of ritual can operate within a short interaction, and which depend on familiarity or repetition over time?

This week, I completed training modules covering HTML webpage building, eye-tracking setup, and understanding eye-tracking metrics. The HTML training helped me think about how to structure the web interface and organize the different stages of the experience. The eye-tracking training helped me consider what data might be available and how I could interpret it when designing the interaction. I still need to decide which measurements would meaningfully support the project’s research questions. I plan to meet with my potential committee members next week, so I do not yet have a committee meeting discussion to report. In those meetings, I hope to discuss the project’s scope, the role of the AI agent, and how to connect the creative experience with a manageable experimental design.

Reference

Carey, J. W. (2009). Communication as culture: Essays on media and society (Rev. ed.). Routledge. https://www.routledge.com/Communication-as-Culture-Revised-Edition-Essays-on-Media-and-Society/Carey/p/book/9780415989763

#5 When an Interaction Becomes a Ritual

In “The Psychology of Rituals: An Integrative Review and Process-Based Framework,” Hobson and colleagues (2018) review research on how rituals influence people’s emotions, performance, and social connections. They describe rituals as structured sequences of actions that carry symbolic meaning and include elements without an obvious practical connection to their intended outcome. Their framework distinguishes between “bottom-up” processes (emerge from performing physical actions) and “top-down” processes (which involve expectations and interpretations). For example, repeating a ordered sequence can direct attention, while the meaning attached to that sequence itself can amplify its significant. The article provides a framework for understanding ritual’s psychological effects and reinds us that ritual do not always produces the same response.

This framework helps me think more carefully about the opening of my installation. I have been considering asking visitors a personal question, guiding their breathing, and inviting them to interact with a machine through their gaze. These actions could establish expectations before the machine performs its apparent mind-reading. For example, following a moving light could be presented as eye-tracking calibration or as a way of establishing a connection with the machine. I am interested in how the same action might feel different depending on its timing, language, and presentation. This gives me a more specific design question: how does an ordinary interface instruction become a meaningful ritual? I want to develop the physical sequence and the accompanying script together, paying attention to pauses, repetition, and the visitor’s role. If I later study participants’ responses, I would also want to distinguish between feeling calm, feeling personally understood, and believing that the machine has special knowledge, and also ask me to think about: how to manipulate those variates in former interaction design. A change in physiological signals alone would not tell me which interpretation a visitor experienced.

This week, I made limited progress, but I browsed examples of related projects and began drafting the visual design for my installation. I mainly considered what kinds of data I could collect and how I could process or transform them within the experience. I am still deciding whether eye-tracking or other physiological signals would best support the interaction. This also made me think about how the interface could turn a bodily signal into something that feels like a personal reading. Next week, I plan to refine the visual draft and map out one possible interaction, identifying what data it would need, how the system would interpret that data, and what the visitor would see.

Reference

Hobson, N. M., Schroeder, J., Risen, J. L., Xygalatas, D., & Inzlicht, M. (2018). The psychology of rituals: An integrative review and process-based framework. Personality and Social Psychology Review, 22(3), 260–284. https://doi.org/10.1177/1088868317734944

#4 When a Machine Feels Like Someone

In “Machines and Mindlessness: Social Responses to Computers,” Nass and Moon (2000) review experiments showing that people apply familiar social rules to computers, even they recognize that computers are not human. These responses include politeness, reciprocity, and judgments based on a computer’s apparent personality or social identity. For example, participants evaluated a computer more positively when that same computer asked about its performance than when they answered through another computer or a paper questionnaire. The authors explain these responses through “mindlessness”: certain cues activate learned social expectations without people fully reconsidering whether those expectations fit the situation. Their argument suggests that a person’s explicit understanding of a machine can differ from how they behave toward it during an interaction.

This distinction is useful for my design because I am interested in the gap between knowing that a system is programmed and feeling that it understands something personal. A visitor might recognize the display as a machine while still responding emotionally to its questions, pauses, and apparent attention. For example, opening with “What question brings you here?” and inviting the visitor to breathe could establish expectations of a personal encounter before the main magic performance begins. This gives me a more specific design question: which parts of that experience come from physiological measurement, and which come from the way the machine addresses the visitor? I could explore this by presenting the same underlying interaction through different scripts, such as a neutral technical explanation and a ritual-like introduction. The article does not establish that these techniques will produce spiritual belief, but it gives me a framework for examining social responses without assuming that visitors literally believe the machine can read minds.

This week, my project work focused on planning the technical skills I need to develop. I identified five hours and forty-nine minutes of training covering Arduino foundations, analog sensors, HTML Canvas graphics, and eye-tracking research. I am still considering whether the first version should center on a browser-based interaction or include a separate physical device. Next week, I plan to sketch a short sequence consisting of an opening question, guided breathing, and a simple choice-based “reading.” I also want to begin the eye-tracking training and identify what equipment and software I could realistically use. This should help me choose one manageable interaction to prototype before adding more sensors or mechanical elements.

Reference

Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81–103. https://doi.org/10.1111/0022-4537.00153

#3 Media Review: Clickclickclick.click and the Magic of Being Watched

I found a website project: Clickclickclick.click (Moniker & VPRO Medialab, 2016), an interactive website with almost no visible content. When you arrive, a plain page shows a single green button. As you move your mouse, scroll, resize the window, or leave the tab, a disembodied voice narrates each action aloud (“The user has moved their cursor to the left,” “The user is hesitating”), and a growing log of “achievements” records what it has observed. The site never asks permission and never explains itself, yet it feels like it looking at you and maybe teasing you (and you are called subject somehow made you feel like you are the laboratory mouse).

What makes the piece work is not technical sophistication but framing. Nothing it detects is secret; it uses ordinary browser events. The magic comes from the voice, the timing, and the escalating sense of a presence on the other side of the screen. This is precisely the effect I want in my project, where a meditative “connect with me” intro leads into a mind-reading act powered by webcam eye tracking. Where Clickclickclick.click is cold and surveillant, I want a warmer, theatrical register closer to a mentalist’s stage act. Accoriding to Derren Brown describes “”in Tricks of the Mind (Brown, 2006).,audiences will attribute intention and insight to a system that simply reflects their own behaviour back at them.

Studying it, the skills I need become clear. Technically, I need front-end development (JavaScript, React, animation libraries such as Rive or GSAP), plus experience integrating browser-based gaze estimation. WebGazer.js (Papoutsaki et al., 2016) is the most accessible option, but its accuracy is limited, so I will need to design “tricks” around dwell time and coarse regions rather than precise coordinates, and build silent (alternative)fallbacks (mouse hesitation, timing) for when tracking fails. Visually, I need illustration and frame-by-frame animation skills for a hand-drawn UI, and sound design for the trance-like intro.

For research, the psychology of cold reading and the Barnum effect, so the “readings” land regardless of tracking accuracy; the ethics and UX of webcam access, including how to secure consent without breaking the illusion; and interactive narrative design, so the experience has a satisfying arc from invitation to reveal. Clickclickclick.click deliberately withholds its reveal; I want to decide whether mine should end with a moment of explanation, turning surveillance-anxiety into wonder rather than unease.

This week’s progress. This week I focused on scoping rather than building. I wrote out the user journey (question prompt, breathing sequence, gaze calibration disguised as ritual, the reading, and a reveal) and researched the tool stack, settling provisionally on React with WebGazer.js for gaze tracking, Rive for hand-drawn interactive animation, and browser-exposed device data (time, battery, timezone, referrer) for the “did my data leak?” moment. I have not written code yet. For next 2 weeks, I plan to build a bare prototype of the calibration-to-card-choice sequence to test how reliably WebGazer can distinguish which of four cards a user is looking at, and to start sketching the visual language for the intro screens.

References

Brown, D. (2006). Tricks of the mind. Channel 4 Books.

Moniker, & VPRO Medialab. (2016). Clickclickclick.click [Interactive website]. https://clickclickclick.click

Papoutsaki, A., Sangkloy, P., Laskey, J., Daskalova, N., Huang, J., & Hays, J. (2016). WebGazer: Scalable webcam eye tracking using user interactions. In Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI 2016) (pp. 3839–3845). AAAI Press.