Friday, December 15, 2017

Teaching A.I. to explain itself

An awkward feature of the artificial intelligence, or machine learning, algorithms that teach themselves to translate languages, analyze X-ray images and mortgage loans, judge probability of behaviors from faces, etc., is that we are unable to discern exactly what they are doing as they perform these functions. How can we we trust these machine unless they can explain themselves? This issue is the subject of an interesting piece by Cliff Kuang. A few clips from the article:
Instead of certainty and cause, A.I. works off probability and correlation. And yet A.I. must nonetheless conform to the society we’ve built — one in which decisions require explanations, whether in a court of law, in the way a business is run or in the advice our doctors give us. The disconnect between how we make decisions and how machines make them, and the fact that machines are making more and more decisions for us, has birthed a new push for transparency and a field of research called explainable A.I., or X.A.I. Its goal is to make machines able to account for the things they learn, in ways that we can understand.
A decade in the making, the European Union’s General Data Protection Regulation finally goes into effect in May 2018. It’s a sprawling, many-tentacled piece of legislation whose opening lines declare that the protection of personal data is a universal human right. Among its hundreds of provisions, two seem aimed squarely at where machine learning has already been deployed and how it’s likely to evolve. Google and Facebook are most directly threatened by Article 21, which affords anyone the right to opt out of personally tailored ads. The next article then confronts machine learning head on, limning a so-called right to explanation: E.U. citizens can contest “legal or similarly significant” decisions made by algorithms and appeal for human intervention. Taken together, Articles 21 and 22 introduce the principle that people are owed agency and understanding when they’re faced by machine-made decisions.
To create a neural net that can reveal its inner workings...researchers...are pursuing a number of different paths. Some of these are technically ingenious — for example, designing new kinds of deep neural networks made up of smaller, more easily understood modules, which can fit together like Legos to accomplish complex tasks. Others involve psychological insight: One team at Rutgers is designing a deep neural network that, once it makes a decision, can then sift through its data set to find the example that best demonstrates why it made that decision. (The idea is partly inspired by psychological studies of real-life experts like firefighters, who don’t clock in for a shift thinking, These are the 12 rules for fighting fires; when they see a fire before them, they compare it with ones they’ve seen before and act accordingly.) Perhaps the most ambitious of the dozen different projects are those that seek to bolt new explanatory capabilities onto existing deep neural networks. Imagine giving your pet dog the power of speech, so that it might finally explain what’s so interesting about squirrels. Or, as Trevor Darrell, a lead investigator on one of those teams, sums it up, “The solution to explainable A.I. is more A.I.”
... a novel idea for letting an A.I. teach itself how to describe the contents of a picture...create two deep neural networks: one dedicated to image recognition and another to translating languages. ...they lashed these two together and fed them thousands of images that had captions attached to them. As the first network learned to recognize the objects in a picture, the second simply watched what was happening in the first, then learned to associate certain words with the activity it saw. Working together, the two networks could identify the features of each picture, then label them. Soon after, Darrell was presenting some different work to a group of computer scientists when someone in the audience raised a hand, complaining that the techniques he was describing would never be explainable. Darrell, without a second thought, said, Sure — but you could make it explainable by once again lashing two deep neural networks together, one to do the task and one to describe it.

Thursday, December 14, 2017

Debussy La plus que lente - a first musical offering from Austin Texas.

This is a personal post, a musical offering of the sort I have done on MindBlog in previous years. The Steinway B that I have used since 2002 recently moved with me from Fort Lauderdale, Florida to Austin, Texas, not to the family house I moved back into, but to the larger living room of my son's home which can manage the kind of musical socials I have given for many years.  Techie MindBlog readers might be interested in my discovery that the video camera on my iPhone X is better than the Canon video camera I had been using, and that a small USB Zoom iQ6 condenser microphone attached to its Lightning connector gives audio quality comparable to the much larger C1 Studio condenser microphone whose output had to be tediously synchronized with video from the Canon camera stripped of its inferior audio sound track.





Wednesday, December 13, 2017

Digital mass persuasion via psychological targeting.

Sigh...we're heading full-tilt towards a plutocracy which will manipulate the masses via technologies of the sort described by Matz et al.:

Significance
Building on recent advancements in the assessment of psychological traits from digital footprints, this paper demonstrates the effectiveness of psychological mass persuasion—that is, the adaptation of persuasive appeals to the psychological characteristics of large groups of individuals with the goal of influencing their behavior. On the one hand, this form of psychological mass persuasion could be used to help people make better decisions and lead healthier and happier lives. On the other hand, it could be used to covertly exploit weaknesses in their character and persuade them to take action against their own best interest, highlighting the potential need for policy interventions.
Abstract
People are exposed to persuasive communication across many different contexts: Governments, companies, and political parties use persuasive appeals to encourage people to eat healthier, purchase a particular product, or vote for a specific candidate. Laboratory studies show that such persuasive appeals are more effective in influencing behavior when they are tailored to individuals’ unique psychological characteristics. However, the investigation of large-scale psychological persuasion in the real world has been hindered by the questionnaire-based nature of psychological assessment. Recent research, however, shows that people’s psychological characteristics can be accurately predicted from their digital footprints, such as their Facebook Likes or Tweets. Capitalizing on this form of psychological assessment from digital footprints, we test the effects of psychological persuasion on people’s actual behavior in an ecologically valid setting. In three field experiments that reached over 3.5 million individuals with psychologically tailored advertising, we find that matching the content of persuasive appeals to individuals’ psychological characteristics significantly altered their behavior as measured by clicks and purchases. Persuasive appeals that were matched to people’s extraversion or openness-to-experience level resulted in up to 40% more clicks and up to 50% more purchases than their mismatching or unpersonalized counterparts. Our findings suggest that the application of psychological targeting makes it possible to influence the behavior of large groups of people by tailoring persuasive appeals to the psychological needs of the target audiences. We discuss both the potential benefits of this method for helping individuals make better decisions and the potential pitfalls related to manipulation and privacy.

Tuesday, December 12, 2017

Trauma is passed over generations.

Bakalar points to work by Santavirta et al. showing that the daughters of women exposed to childhood trauma are at increased risk for psychiatric disorders. The study compared the health of female offspring of ~47,000 Finnish children who were evacuated to Swedish foster homes during World War II, with offspring of female cousins who had not been evacuated. The study:
...found that female children of mothers who had been evacuated to Sweden were twice as likely to be hospitalized for a psychiatric illness as their female cousins who had not been evacuated, and more than four times as likely to have depression or bipolar disorder...But there was no effect among male children, and no effect among children of either sex born to fathers who had been evacuated.

Monday, December 11, 2017

How is American (and World) governance evolving?

So... what is the United State to become? From the recent outpouring of Op-Ed pieces you can take your choice: Autocracy, Plutocracy, Oligarchy, Kleptocracy... with liberal democracy viewed as vitally threatened. Articles by Thomas Edsall and Andrew Sullivan describe how American democracy is destroying itself, as Roger Cohen sadly notes the irreversible passing of the Pax Americana, an ordering of the world that began with Woodrow Wilson's 14 points speech one hundred years ago. David Frum outlines steps towards Autocracy as Jonathan Rauch discusses whether Trump will be able to govern as an authoritarian. Paul Krugman notes how the current tax reform will enormously enhance the ongoing process of entrenching a hereditary plutocracy that actually runs the country.  Articles by Fareed Zakaria and Thomas Edsall describe how the liberal establishment has failed to understand its own role in the rise of contemporary conservatism, how its social and economic policies have disadvantaged formerly middle class voters more motivated by issues surrounding religion, race, and culture than they are by economics, thus fueling a rise of nationalism, nativism and xenophobia in both the U.S. and Europe.  Regarding this last point, I want to paste in here the final paragraphs of an Edsall Op-Ed piece noting Eric Schnurer's argument that blue America has over the last decade declared war on the "red way of life."
The political, economic, and cultural triumph nationwide of a set of principles and realities essentially alien to large numbers of Americans is viewed as (a) being imposed upon them, and (b) overturning much of what they take for granted in their lives — and I don’t think they’re wrong about that. I think they’ve risen in angry revolt, and now intend to give back to the “elite” in the same terms that they’ve been given to. I don’t think this is good — in fact, I think it’s a very dangerous situation — but I think we need to understand it in order to responsibly address it.
Do liberals in fact need to understand — or empathize with — their many antagonists, the men and women who are sharply critical of the liberal project?
Steven Pinker, a professor of psychology at Harvard, observes that “believers in liberal democracy have unilaterally disarmed in the defense of the institution” by agreeing in many cases with the premise of the Trump campaign: “that the country is a hopeless swamp.” This left Democrats “defenseless when he proposed to drain it.” Where, Pinker asks,
are the liberals who are willing to say that liberal democracy has worked? That environmental regulations have slashed air pollutants while allowing Americans to drive more miles and burn more fuel? That social transfers have reduced poverty rates fivefold? That globalization has allowed Americans to afford more food, clothing, TVs, cars, and air-conditioners? That international organizations have prevented nuclear war, and reduced the rate of death in warfare by 90 percent? That environmental treaties are healing the hole in the ozone layer?
Pinker remains confident:
Progress always must fight headwinds. Human nature doesn’t change, and the appeal of regressive impulses is perennial. The forces of liberalism, modernity, cosmopolitanism, the open society, and Enlightenment values always have to push against our innate tribalism, authoritarianism, and thirst for vengeance. We can even recognize these instincts in ourselves, even in Trump’s cavalier remarks about the rule of law...Over the longer run, I think the forces of modernity prevail — affluence, education, mobility, communication, and generational replacement. Trumpism, like Brexit and European populism, are old men’s movements: support drops off sharply with age.
Pinker is optimistic about the future. I hope he is right.
The problem is that even if Pinker is right, his analysis does not preclude a sustained period in which the anti-democratic right dominates American politics. There is no telling how long it will be before the movement Trump has mobilized will have run its course. Nor can we anticipate — if and when Trumpism does implode — how extensive the damage will be that Pinker’s “forces of modernity” will have to repair.
But... what if all of this wringing of hands about changes the political order is a thin veneer over deeper changes that are really going to end up controlling the show?  One is seeing now the rise of a de facto world government of interlocked and interdependent giant corporations, mainly in the U.S. and China (think Apple and Foxconn) versed in the neuroeconomic techniques central to influencing the behaviors, desires, and consumptions of their subjects.  They are assembling a level of power that might increasingly override the ability of individual nation states to contest or control their actions.  Will this ensemble nudge towards mirroring the values of liberalism currently reflected in the public stances of the largest U.S. corporations, or will the political accommodations shown by  their Asian counterparts be more likely to prevail?

Friday, December 08, 2017

Dogs can smell our happiness and fear.

From D'Aniello et al:
We report a study examining interspecies emotion transfer via body odors (chemosignals). Do human body odors (chemosignals) produced under emotional conditions of happiness and fear provide information that is detectable by pet dogs (Labrador and Golden retrievers)? The odor samples were collected from the axilla of male donors not involved in the main experiment. The experimental setup involved the co-presence of the dog's owner, a stranger and the odor dispenser in a space where the dogs could move freely. There were three odor conditions [fear, happiness, and control (no sweat)] to which the dogs were assigned randomly. The dependent variables were the relevant behaviors of the dogs (e.g., approaching, interacting and gazing) directed to the three targets (owner, stranger, sweat dispenser) aside from the dogs' stress and heart rate indicators. The results indicated with high accuracy that the dogs manifested the predicted behaviors in the three conditions. There were fewer and shorter owner directed behaviors and more stranger directed behaviors when they were in the "happy odor condition" compared to the fear odor and control conditions. In the fear odor condition, they displayed more stressful behaviors. The heart rate data in the control and happy conditions were significantly lower than in the fear condition. Our findings suggest that interspecies emotional communication is facilitated by chemosignals.

Thursday, December 07, 2017

How Evil is Tech?

The title of this post is from an Op-Ed piece by David Brooks. Some clips:
There are three main critiques of big tech.
The first is that it is destroying the young. Social media promises an end to loneliness but actually produces an increase in solitude and an intense awareness of social exclusion. Texting and other technologies give you more control over your social interactions but also lead to thinner interactions and less real engagement with the world.
The second critique of the tech industry is that it is causing this addiction on purpose, to make money. Tech companies understand what causes dopamine surges in the brain and they lace their products with “hijacking techniques” that lure us in and create “compulsion loops.”
The third critique is that Apple, Amazon, Google and Facebook are near monopolies that use their market power to invade the private lives of their users and impose unfair conditions on content creators and smaller competitors. The political assault on this front is gaining steam.
The big breakthrough will come when tech executives clearly acknowledge the central truth: Their technologies are extremely useful for the tasks and pleasures that require shallower forms of consciousness, but they often crowd out and destroy the deeper forms of consciousness people need to thrive...Online is a place for human contact but not intimacy. Online is a place for information but not reflection.
Rabbi Abraham Joshua Heschel wrote that we take a break from the distractions of the world not as a rest to give us more strength to dive back in, but as the climax of living. “The seventh day is a palace in time which we build. It is made of soul, joy and reticence,” he said. By cutting off work and technology we enter a different state of consciousness, a different dimension of time and a different atmosphere, a “mine where the spirit’s precious metal can be found.”
Imagine if instead of claiming to offer us the best things in life, tech merely saw itself as providing efficiency devices. Its innovations can save us time on lower-level tasks so we can get offline and there experience the best things in life.
Imagine if tech pitched itself that way. That would be an amazing show of realism and, especially, humility, which these days is the ultimate and most disruptive technology.

Wednesday, December 06, 2017

Trust and cooperation across societies.

Romano et al. (open source) offer, in a study over 17 countries, an example of the kind of research needed to understand and enhance cooperation within and between groups.

Significance
In a study including 17 societies, we found that people are motivated to trust and cooperate more with their ingroup, than harm the outgroup. Reputation-based indirect reciprocity may offset this ingroup favoritism, because we found that reputational concern universally increases cooperation with both ingroup and outgroup members. We also found that people who are dispositionally cooperative are less parochial and more universal in their cooperation. In a time of increasing parochialism in both domestic and international relations, our findings affirm us of the danger of the strong human universal toward parochial altruism. Yet, our findings suggest that in all societies, there exist people whose cooperation transcends group boundaries and provides a solution to combating parochialism: reputation-based indirect reciprocity.
Abstract
International challenges such as climate change, poverty, and intergroup conflict require countries to cooperate to solve these complex problems. However, the political tide in many countries has shifted inward, with skepticism and reluctance to cooperate with other countries. Thus, cross-societal investigations are needed to test theory about trust and cooperation within and between groups. We conducted an experimental study in 17 countries designed to test several theories that explain why, who, and where people trust and cooperate more with ingroup members, compared with outgroup members. The experiment involved several interactions in the trust game, either as a trustor or trustee. We manipulated partner group membership in the trust game (ingroup, outgroup, or unknown) and if their reputation was at stake during the interaction. In addition to the standard finding that participants trust and cooperate more with ingroup than outgroup members, we obtained findings that reputational concerns play a decisive role for promoting trust and cooperation universally across societies. Furthermore, men discriminated more in favor of their ingroup than women. Individual differences in cooperative preferences, as measured by social value orientation, predicted cooperation with both ingroup and outgroup members. Finally, we did not find support for three theories about the cross-societal conditions that influence the degree of ingroup favoritism observed across societies (e.g., material security, religiosity, and pathogen stress). We discuss the implications for promoting cooperation within and between countries.

Tuesday, December 05, 2017

The emotional political base supporting the consolidation of the U.S. plutocracy.

I want to pass on the ending paragraphs from a piece by Fareed Zakaria:
Is it that the Republican Party is cleverly and successfully hoodwinking its supporters, promising them populism and enacting plutocratic capitalism instead? This view has been a staple of liberal analysis for years, most prominently in Thomas Frank’s book “What’s the Matter with Kansas?” Frank argued that Republicans have been able to work this magic trick by dangling social issues in front of working-class voters, who fall for the bait and lose sight of the fact that they are voting against their own interests. Both Wolf and Pierson believe that this trickery will prove dangerous for Republicans. “The plutocrats are riding on a hungry tiger,” writes Wolf.
But what if people are not being fooled at all? What if people are actually motivated far more deeply by issues surrounding religion, race and culture than they are by economics? There is increasing evidence that Trump’s base supports him because they feel a deep emotional, cultural and class affinity for him. And while the tax bill is analyzed by economists, Trump picks fights with black athletes, retweets misleading anti-Muslim videos and promises not to yield on immigration. Perhaps he knows his base better than we do. In fact, Trump’s populism might not be as unique as it’s made out to be. Polling from Europe suggests that the core issues motivating people to support Brexit or the far-right parties in France and Germany, and even the populist parties of Eastern Europe, are cultural and social.
The most important revolution in economics in the past generation has been the rise of the behavioral scientists, trained in psychology, who are finding that people systematically make decisions that are against their own “interests.” This might be the tip of the iceberg in understanding human motivation. The real story might be that people see their own interests in much more emotional and tribal ways than scholars understand. What if, in the eyes of a large group of Americans, these other issues are the ones for which they will stand up, protest, support politicians and even pay an economic price? What if, for many people, in America and around the world, these are their true interests?

Monday, December 04, 2017

A mind reading machine?

Not quite, but Matthew Hutson points to work by Wen et al. using an artificial neural network to categorize fMRI signals from subjects watching different categories of images. The algorithm could predict with about 50% accuracy which of 15 classes of visual object a subject was watching. His description:
Artificial intelligence has taken us one baby step closer to the mind-reading machines of science fiction. Researchers have developed “deep learning” algorithms—roughly modeled on the human brain—to decipher, you guessed it, the human brain. First, they built a model of how the brain encodes information. As three women spent hours viewing hundreds of short videos, a functional MRI machine measured signals of activity in the visual cortex and elsewhere. A popular type of artificial neural network used for image processing learned to associate video images with brain activity. As the women watched additional clips, the algorithm’s predicted activity correlated with actual activity in a dozen brain regions. It also helped the scientists visualize which features each area of the cortex was processing. Another network decoded neural signals: Based on a participant’s brain activity, it could predict with about 50% accuracy what she was watching (by selecting one of 15 categories including bird, airplane, and exercise). If the network had trained on data from a different woman’s brain, it could still categorize the image with about 25% accuracy, the researchers report this month in Cerebral Cortex. The network could also partially reconstruct what a participant saw, turning brain activity into pixels, but the resulting images were little more than white blobs. The researchers hope their work will lead to the reconstruction of mental imagery, which uses some of the same brain circuits as visual processing. Translating from the mind’s eye into bits could allow people to express vivid thoughts or dreams to computers or to other people without words or mouse clicks, and could help those with strokes who have no other way to communicate.

Friday, December 01, 2017

Bringing big science to psychology

Chawla describes a new initiative dubbed the "Psychological Science Accelerator" (PSA) that:
...has so far forged alliances with more than 170 laboratories on six continents in a bid to enhance the ability of researchers to collect data at multiple sites on a massive scale...to enable researchers to expand their reach and collect “large-scale confirmatory data” at many sites.
A selection committee has evaluated eight proposals and selected one based on experiments already replicated in the US and the UK.
It aims to discover whether the research findings of Alexander Todorov, a psychologist at Princeton University, can be replicated on a global scale. Todorov has reported that people rank human faces on two components: valence and dominance. Valence is a measure of trustworthiness, whereas dominance is a measure of physical strength...More than 50 of PSA’s collaborating labs have already committed to collect data as part of the study.
PSA isn’t the only effort aiming to change how researchers conduct psychological studies, which have received extensive criticism for a lack of reproducibility. Others include the Many Labs Replication Project and the Pipeline Project. Earlier this year, Chartier also launched StudySwap, an online platform designed to help researchers find collaborators for replication studies and exchange resources.