Robopsychologie

• The brain has developed a "rulebook" of how objects should look from past experiences
• When interpreting new information, our brain relies on previous knowledge
• the brain takes a "shortcut" to focus on important aspects
• optical illusions fool our brains by taking advantage of these shortcuts

• 9-month-old infants were able to pass rapid categorization of human and ape faces (Peykarjou et al., 2017)

• Perceptual categorization of cat and dog silhouettes by 3- to 4-month-old infants (Quinn, Eimas & Tarr, 2001)

• Generative Adversarial Networks (GANs) revealed a new challenge for image classification: Adversarial Images
• Adversarial Images are images whose class category looks obvious to a human but causes massive failures in a deep network
• with only a minor disortion (seemingly) a deep network's classification of the image goes from a panda to a gibbon

• It is necessary to highlight the fundamentally flawed ways that ImageNet classifies people in “problematic” and “offensive” ways.

• It is crucial to assess the fallibility of AI systems and prevalence of machine learning bias

• Given a single facial image, a classifier could correctly distinguish between gay and heterosexual men in 81% of cases, and in 71% of cases for women

• Human judges achieved much lower accuracy: 61% for men and 54% for women

• The accuracy of the algorithm increased to 91% and 83%, respectively, given five facial images per person.

• Facial features employed by the classifier included both fixed (e.g., nose shape) and transient facial features (e.g., grooming style).

• Is it important for users to be able to distinguish between human and machine (e.g. on the phone)? Would you like to know?

• Are artificial voices, which sound very humanlike, sometimes creepy?

• Would you prefer to talk to a realistically human sounding bot or one that clearly sounds like a machine?

• Are these preferences perhaps context-dependent?

• Which visual image of a virtual conversation partner does a more or less human-like voice actually evoke?

• Human-likeness of the drawn robots was generally high across all conditions.

• Some features appeared in almost all drawings regardless of the voice (e.g., head, eyes).

• Other features were significantly more prevalent in voice conditions characterized by low human-likeness (wheels) or high human-likeness (e.g., nose).

• Some studies suggest that female synthetic voices are preferred (voicebot.ai, 2019) as they are perceived as warmer compared to male voices (Karl MacDorman, Indiana University).

• Other studies revealed the opposite:

  • Results indicated that female human speech was rated as preferable to female synthetic speech,

    and that male synthetic speech was rated as preferable to female synthetic speech

    (Mullenix et al., 2003).

  • Male voices are perceived as more intelligent (Clifford Nass, Stanford University)

• Between 145-175 Hz (= gender-neutral range according to research)

• Voice was refined after surveying 4,600 people

• Collaboration: Copenhagen Pride, Virtue, Equal AI, Koalition Interactive & thirtysoundsgood

1. Encoding (Acquiring the memory and consolidating it)
2. Storage (How a memory is maintained (aufrechterhalten)
3. Retrieval (How a memory is retrieved, (erneuert)

Short Term Memory:

1. Capacity is limited
2. Quick to learn
3. Quick to forget
4. 7+2 capacity

Long Term Memory

1. Practically unlimited
2. slow to learn
3. slow to forget

Frontal lobes (Short term memory tasks)

Prefrontal Cortex, Parts of temporal lobes (Efficient encoding of words, pictures)

Hippocampus (Formation of long-term declarative memories, may bind together diverse elements of a memory so it can be retreived later as a coherent etntity

Cerebellum (Formation an retention of simple classically conditioned responses

Cerebral Cortex (Storage of long-term memories, possibily in areas involved in the original perception of the information)

• Most adults have few memories of events below age 3

• And from 3-7 years they have fewer memories than would be predicted by forgetting alone

• However, studies have shown that 3-year-olds can form episodic memories (Fivush et al 1987, Hamond & Fivush, 1991, Sheingold & Tenney, 1979)

1. Explanation (Bauer et al., 2007):
   • In adults, power function à over time, forgetting slows, presumably (vermutlich) as a result of consolidation
   • In children, exponential function -> forgetting continues at a constant rate

◦ 8 for college students ◦ 6-7 for 12-year-olds
◦ 4 for 5-year-olds

The average person‘s short-term memory can remember 7 numbers.

Computing wins:
Input and output
Information processing and memory

Closely Matched: 
Complex movement
Vision
Language
Structured Problem solving

Brain Wins: 
Creativity
Emotion and empathy
Planning and executive function
Consciousness

• But: Humans are able to combine the different memory types, learn from their past experiences and draw conclusions for upcoming experiences

Memory for Faces

• Own-race bias in memory for faces?

• Meta-analyis (Meissner & Brigham, 2001):

◦ Overall, results indicated a "mirror effect“ pattern in which own-race faces yielded a higher proportion of hits and a lower proportion of false alarms compared with other-race faces.

◦ Could be explained with perceptual learning (see perception lecture).

• The so-called sleeper effect means that after a certain period of time recipients might only remember the content of a message, but forget the source and how trustworthy the source was.

• This means: Even if recipients identify fake news as such and therefore initially classify them as not trustworthy, it can happen that after a few weeks they only remember the content of the report but forget that they originally thought it was untrue. They can no longer attribute the information to any source.

Mechera-Ostrivsky & Gluth (2018): Memory Beliefs Drive the Memory Bias on Value- based Decisions

• For value-based decisions, people need to retrieve relevant information from their memory

• Memory bias arises because people believe they remember better options more often than worse options

• In their study, memory performance was higher for more attractive options = letting decisions be influenced by memory could be an adaptive strategy?

• However, the memory bias also persisted after correcting for the effect = this suggests that it is not simply an artifact of unequal memory performance

◦ Their study showed that language was able to alter participants memory of a car accident

◦ Depending on how the questions were frased, participants would remember the car‘s speed differently

The question was:

“About how fast were the cars going when they (smashed / collided / bumped / hit / contacted) each other?”

• The results revealed that the estimated speed varied depending on the verbs used.

• The verb affected the participant‘s memory of the event.

• Since a crime is an unplanned event and often emotional & stressful event = eye witnesses are not prepared and unable to focus on smaller details

• Race bias, previous experiences, knowledge, clothing etc. can all play a role in making the wrong decision as an eye witness

◦ Dementia (e.g Alzheimer‘s disease)
◦ Amnesia, head injury, stroke
◦ Stress, Drug use (also alcohol & smoking)
◦ Nutritional deficiency (Ernährungsmangel)
◦ Sleep deprivation

Incorrectly remembering one‘s past attitudes and behaviour as resembling present attitudes and behaviour

Seeing new information as confirming your established opinion

• Priming means to prepare = so a priming effect happens in demonstration where a preceding stimulus influences the processing or perception of a subsequent stimulus
• Different types of priming: Semantic Priming, conceptual, negative, positive, affective etc.
• Priming enhances retrieval of information and is cloesely linked to implicit memory

Classical Conditioning (Associate an involuntary response and a stimulus)
Famous study by the Russian physiologist Ivan Pavlov and his dogs (Pavlov‘s dogs)

• His classic conditioning study focuses on an association between two stimuli –
• a pre-existing stimuli (unconditioned)
• a neutral one
  For Pavlov‘s dogs the smell of food naturally triggered them to salivate = unconditioned response
• Conditioning the sound of the bell to be related to the smell of food and thus salivation = the dogs will learn to salivate when they hear the sound of the bell (without the smell of food)

Operant Conditioning (B.F. Skinner)

• This type of conditioning focuses on punishment or reinforcement learning

• This type of conditioning can create an association between the consequence of a behaviour and results in decreasing or increasing a particular type of behaviour

• If a pupil is being loud in class and as a result is being punished by doing extra school work – the pupil can learn to build an association between the behaviour (being loud in class) and the consequence (being punished with extra work) which will result in a decrease of this behaviour over time.

• Operant Conditioning (Associate a voluntary beaviour and a consequence)

• Putting human experience and need at the center of technological development
• Applying scientific methods and theories in psychology to current questions of artificial intelligence
  • practical direction: Creating evidence-based guidelines for human-centered technology, collaborating with other disciplines & industry

Agent specific factors

visual form, behavior, communication design

User specific factors

age, personality, technical experience

Contextual factors

area of application, information provided

• The tipping rate of both male and female customers was found to be significantly higher in both of the touching conditions than in the no-touch condition.

• Recent studies show similar effects: Both men and women who are lightly touched by a woman on the back are more likely to take bigger financial risk in an investment game than those not touched at all, or touched by a man (e.g., Levav & Argo, 2010).

• 76 Indiana University students, none of whom were black, were greeted in the research lab by an African-American woman who wore a Black History Month T-shirt “to heighten participants’ awareness of her racial identity.” Each participant sat down in front of a computer. Half of them were given a light touch on the shoulder for one to two seconds.

• All performed an implicit association test to measure unconscious racism. They had to categorize a series of 144 words, which flashed briefly onto the computer screen, as either good or bad. (Example: “pleasant” is good, “disaster” is bad.) Photographs of faces—black and white—popped up between the words. Taking longer to correctly identify a good word immediately after seeing a black face (or, conversely, moving especially quickly to identify a bad word) indicates implicit racism.

• Those who had been touched by the black experimenter were shown to have significantly more positive implicit attitudes toward blacks.

• The topic is still new and results are mixed. Various findings suggest that a touch by a robot can indeed enhance positive feelings towards the robot (e.g. Fukuda et al, 2012, on the „Midas Touch in Human-Robot Interaction“), reduce stress and increased the perceived intimacy of the human–robot bond (Willemse & Erp, 2019) or increase compliance (Hoffmann, 2017).

• Others suggest that an „empathic touch“ by a robot is not necessarily perceived positively (Chen et al., 2011).

• People responded significantly more favorably to the instrumental touch than to the supposedly affective touch.