05 - Non-Verbal Com, Gaze, Controlled Experiments Flashcards
What is non-verbal interaction?
- everything not involving words or speech
- gaze, face expression, gestures, posture, etc
- Channels include:body language, distance, voice (speed, pitch etc), touch
Why is it important in HRI/ how can robots benefit from it
- a lot of human interaction happens non-verbally
- better understanding of whats between the lines
- Robots created for interaction but not understanding NVC signals and or lacking NVC signals are rated worse
- “Designing HRI that meets social norms and cultural expectations might mean the difference between a successful product and a wasted investment.” – HRI book
In what situations could there be a lack of NVC?
- Online communication, phone calls etc
- Autism
- Intercultural misunderstanding
Name different types of gestures and their impact.
- Types:
- Deictic – pointing out objects in an environment
- Iconic – supporting, illustrating speech
- Symbolic – carry their own meaning, even without speech (wave hello, bye)
- Beat gestures – emphasizing rythm
- Impact:
- Matching appropriate gesture with speech leads to wider acceptance
- Mismatching leads to wider rejection of such robots
What is the difference between mimicry and imitation?
- mimicry: unconscious replication
- imitation: conscious replication
Explain the importance of riming and rhytm in HRI.
Timing:
- Turn taking in verbal interaction
- Nonverbal cues can support turn taking → eg looking at a person when you are done speaking and expect an answer
Rhythm:
- Communication between people follows a rhythm
- If the robot is out of sync, it affects people’s opinion about it
- To move and speak at the right time to enable smooth communication
- Entrainment –synchronization of people to an external perceived rythm
Name and explain different eye gaze actions between agents
Mutual gaze
- agents looking at each other and become aware of this → eye contact
Joint Attention
- two or more agents pay visual attention to an object with all sides being aware of this
Gaze aversion
- an agent averts its gaze to signal thinking
Selection by gaze
- an agent looks at a person with the intention to select that person for interaction
What is the usual scientific method procedure?
- Recognize Problem
- Study literature
- Formulate hypothesis
- Perform Experiment
- Analyze Data
- Hypothesis supported
- If not, loop back to 2
What is a hypothesis?
A scientific predicition of outcome, a statement of what is expected to happen.
It has to be objective, testable, falsifiable (the opposite needs to be able to be proven), original
What are independent and dependent variables?
Independent Variables:
- An input that we can change to cause effects, eg. robot emotions
- Usually categorical: happy, neural, sad etc
- we want to control these
Dependent Variables:
- Output variables that we measure to quantify our output
- Continuous → eg. reaction time
- Categorical → eg. likert scale
What is the difference between Between Subjects and Within Subjects Setup?
Between Subjects Setup
- Subjects of experiment are assigned to different conditions
- E.g.placebo testing of a new medication
Within Subjects
- Each subject completes all levels
- E.g.one person interacts with the robot in happy, neutral and sad conditions
- Counterbalancing
- Pro: much more data/less people
- Con: Interdependence between the levels
What are nuisance and confounding factors?
Nuisance Factors
- The annoying ones
- Have effect on the outcome, but not interesting for experimenter
- Need to be minimized, explained, accounted for
- Gender, age, environmental differences, etc.
Confounding Factors
- The scary ones
- Influences both the independent and the dependent variable
- Can be a hidden independent variable
- Example: independent variable: activity level, dependent variable: weight gain, confounding variable: age
What is the difference between accuracy and precision?
Accuracy: closeness of the measurements to a specific value
Precision: closeness of the measurements to each other
What is the p-value?
- how high is the probability for obtaining these results if the null hypothesis is correct
- if p<0.05 → samples come from two difference populations
- if p>0.05 we cannot be sure
(in theory a different alpha value can be chosen)
Name and explain different scales of variables.
Nominal: named categories with no natural order (gender, major, eye color). Can be binary (eg. yes/no)
Ordinal: Name categories, with natural order but difference values might not be consistent
(education level, income, turnament rank). Appropriate to measure: median and quantiles, Inappropriate: mean
Continuous - interval:
order of values and the interval/distance between any two points is meaningful
- Appropriate to measure: mean, standard deviation
- no zero measurement for lack of characteristic
Continuous - Ratio: with 0 value for lack of measurement
What is the difference bwtween descriptive and inferential statistics?
Descriptive:
- does not assume properties of population, just summarized/visualizes sample
- mean, standart deviation, variance
Inferential:
- makes conclusions about whole population from sample data
- hypothesis test, confidence intervals, regression analysis, ANOVA, T-test, etc
What is regression used for?
Estimation of the relationship between variables -> how the typical (median) value of the dependent variable changes when any of the independent variables vary
What is the independent T-Test used for?
Compares means between two unrelated groups on same dependent variable
(1 independent variable, 2 levels, between subjects)
Non-parametric equivalent: Mann-Whitney
What is the dependent T-Test ?
- paired samples t-test
- same participants are tested more than once
- compares means of two related groups on same continuous dependent variable
(1 independent variable, 2 levels, within subjects)
Non-parametric equivalent: Wilcoxon
Explain One-Way, Repeated Measures and Two-Way ANOVA.
ANOVA
→ ANalysis Of Variance
- statistical difference between means of three or more groups (levels)
- multiple t-test would increase the type 1 error
One Way:
1 independent variable, 3+levels, between subjects
Non-parametric equivalent: Kruskal-Wallis
Repeated Measures:
1 independent variable, 3+levels, within subjects
Non-parametric equivalent: Friedman
Two-Way:
2 independent variable
Non-parametric equivalent: Two-Way Friedman
What is the central limit theorem?
What types of biases are there?
Cognitive
- anchoring (reliance on first information)
- attribution (Own actions results of external factors; others’ actions reflection of internal factors)
- confirmation (Search for information that confirms one’s beliefs, neglect information which contradicts it)
- halo (Overall impression of a person affects specific perception)
- self-serving (Cognition is affected by the need of a person ho enhance their self-esteem)
- status quo (Current baseline is taken as a reference point)
Contextual
- academic (Bias of scholars allowing their beliefs to shape their research)
- experimenter bias (Experimenter’s expectations affect the outcome of the experiment)
- funding bias (Outcome of experiment tend to support the funding agency’s views –wonder why)
- publication bias (What is publishable shapes research)
Statistical
- observer-expectancy effect (A researcher’s expectations causes them to inadvertently influence the results)
- reporting (Selective revealing or suppression of information to one’s own gain)
- social desirability (Survey participants will answer questions in a way that is more socially acceptable)
- selection (Conscious or subconscious selection of data which is not representative)
- survivorship