6 Challenges of the future of Reputation and Trust
I would like to share some of the main challenges we have been working on at Traity for the last few months. Building solid trust and reputation systems for the sharing economy is difficult, first because there is not enough research around it, and secondly because many of the challenges come from moral and ethical points of view rather than simply technical.
We have been working with Universities, experts and thought leaders, and here are some of the things I have learnt. At the end of the post I raise 6 debates to open a discussion for the future.
To understand reputation, what it is, its past, present, and future, I shall define two concepts first:
Closed networks and Open networks
- In a closed network, everybody knows everybody else. If you become a bad actor (and become the red circle on the left), the rest of us will sooner or later learn about this and will exclude you from the network.
- In an open network, you have people who don´t know each other. A bad actor can behave poorly on one branch and keep the other network of friends intact.
Reputation exists to regulate social interactions, to create social control, aiming to eliminate bad actors.
In the past, in small villages, most interactions happened as part of closed networks. The village would have a few hundred people and most buyers and sellers would know each other, so we would have a previous history of buying our cheese from the same people. It was unlikely that we would ever get low quality cheese from those sellers, because their reputation would be damaged in front of all the other villagers, so the incentives to keep high quality were high.
This is also the reason why it was difficult to move villages in the past. As a new external actor, you would have to prove your reputation from zero with hard work until you had a sufficiently large history of positive transactions in the new village, or prove your reputation by bringing a letter of recommendation from a known source of the previous village (but in many cases this would be hard to prove, particularly when the villages were in different countries. If you read the book “The Physician” there are many interesting stories that have to do with reputation of a villager who travels from city to city from London to Iran, and how he deals with those).
Today, we still have closed networks in groups of friends where every friend knows everybody else and you have an incentive to be a good actor with all of them, but as we live in large cities, in most cases we don´t know all the buyers and sellers of the market, so our networks are often open.
Furthermore, today we travel a lot more and move from city to city almost seamlessly. There are a few reasons why this sociological changes have worked out: You can prove your previous history with CVs that are validated by the employers, and as organizations become large and known, this serves as a “trusted source”. If somebody today can prove that they worked for Google or that they have a degree from Harvard, the endorsement from that institution opens the doors in the new cities. You are not trusting the person, you are trusting the institution. It’s a “Third party reputation system”. Or if you buy the cheese at large supermarkets, the whole supermarket brand is putting its reputation at risk, so the quality of your cheese is ensured by the strength of the brand. Basically, reputation today is a reputation of trusting brands as third parties.
The Future: Let´s say you want to hire the best pet detective, and a scruffy little guy shows up. As explained before, there are two ways he can prove his reputation to you:
- Previous history of transactions, a track record.
- Endorsement from a trusted source, somebody you already know who is trusted. This can be an institution: The Harvard of Pet Detectives
Unfortunately it will not be easy for Ace Ventura to prove neither his track record nor an endorsement from a trusted source. Firstly, because his CV is full of gigs here and there that are difficult to prove, and secondly, because there is no Harvard of Pet Detectives, no institutional endorser.
In the future we will increasingly work with people like Ace. Freelancers who deliver products and services that you value. We will need to trust those people, who might come without a brand behind. Arun Sundararajan, professor at New York Stern University, suggested recently that the freelance workforce will become 50% of the population in the coming years.
How can we build solid reputation systems so that Ace and many others can prove their reputation seamlessly and enable those workers to perform their work effectively?
Let’s take a look at what happens in the digital world with Amazon. We have learned to trust reviews of books to determine their quality and to help us make decisions about them. This is an interesting phenomena, because we don´t know any of the reviewers, but here we are not putting our trust on the reviewers, we are putting our trust on the statistical significance of the reviews. If there are 70 reviews with 4.5 stars, that´s better than 1 review with 5 stars.
Unfortunately (or fortunately) we cannot really do this to review a person effectively. Humans are complex and cannot be reduced to a number between 1 and 5 “as a human”.
We have eBay, where we are measuring people’s abilities to send a package flawlessly. Or AirBnB where we measure people’s cleanness or ability to be at places on time, or BlaBlaCar, where we are measuring people’s driving skills. All of these are different traits of how people behave on different transactional businesses. Together with traditional data, such as institutions you have worked at as endorsers, your presence in communities like non profits, the reach of your social network, your online presence like blogging, and other proven recommendations of your freelance work on specialized sites, like Dribbble for designers, will help people determine your reputation and whether you are the right person to perform some work.
One of the main characteristics of this is the ability to extrapolate reputation, as we do more and more different types of jobs. If you want to start a new freelance gig where you look after people’s Iguanas because you had one for many years and know how to do it, people might need to see that you have a previous history looking after people’s iguanas. But of course if you have just signed up to “shareiguanas.com”, you will have no previous reviews. Here is where extrapolation will become important.
- Firstly, you will have to prove your experience with your own iguana, possibly sharing pictures of your own iguana, or having blogged about it in the past.
- Secondly, you will want to prove other traits that are important in looking after other people´s iguanas. This might include, for example, whether you arrive to places on time. Since this trait is the same as the trait needed for AirBnB, you should be able to use such previous history of successfully being on time at places on AirBnB to help you land the Iguana pet sitting job. You can argue the same about being clean or being friendly, which all are important traits to work with someone, and you can prove those traits from other interactions in diverse environments where the same traits are important and are proven.
These are some of the trends I believe will become part of your reputation in the future:
- Your history of transactions will be infinite, and will include all the transactions that you did in the past, in very diverse settings, each of them adding to your reputation in different traits that might be relevant or not for the next task.
- Endorsement will be anyone to anyone in a multi-layer chain of contact. Moving from just “brand endorsement” to “weak links of reputation” (+1s) for each interaction we have with different people. This will enable massively open and liquid ephimeral networks.
- Offline-to-Online Reputation will grow in importance: my mother doesn’t use eBay or Airbnb, but she has a good reputation in her circles. How can we be inclusive to bring people like my mother on to the “economic graph” of online opportunities? We will have to find ways to recommend her offline that prove her reputation online. I envision this as the biggest challenge for the next 10 years, because when you think about it, of the 9 billion people in the world, only 1 million people have reviews on AirBnB.
- Each weak link is contextual, adding to the traits that are relevant for each of those weak links. The extrapolation I mentioned earlier.
- The negative links have a lot of value. When you see someone on eBay with 78 positive ratings and 1 negative rating, what do you do next? check what happened with that negative rating. Why? Because that´s the only piece of information of the total 79 reviews that really matters. This is a critical aspect to consider.
Which leads to some of the unsolved debates that need to be defined in the coming years:
1. Who owns your reputation?
Reputation, semantically, is defined as “what other people think about you”. So it is difficult to put ownership on it. Do you own the opinions? Or do you own where those opinions are stored?
Online, eBay owns your eBay reputation. AirBnB owns your AirBnB reputation. Perhaps you should own your own reputation instead, in such way that you can leverage it in any way you want, in order to look after people´s iguanas.
2. Right to forget and Right to delete?
However, based on point 1, should you be able to do anything about those opinions? Are you the owner of those opinions or is the reviewer the owner? If you are the owner of your reputation, should you be able to delete them? Or modify them? . We just mentioned that the most valuable links are the negative links. If we delete them, what’s the point? Reputation is worthless, I can’t trust anyone.
My opinion here is that, as many transactions become public (like those on eBay or AirBnB, for everything else) we probably need to look at the mixed model, where people cannot simply delete opinions, but older opinions would have less weight, or after a few years, get completely forgotten. Maybe things that happen earlier than the last 5 years should not affect your reputation. You were a rebel teenager who defrauded someone on eBay for $70 and now you can prove that you are a good citizen.
3. The value of negative reputation is High.
As discussed earlier, if an eBay seller has 99.7% positive reviews, you check the 0.3% of negative reviews. Negative reputation data is critical for decision making, but how much is good and how much is bad? Is 99.7% good? Is 95% good? Is 90% good? Is 70% good? Where is “good”. It is probably a subjective metric. Might threshold might be on 99.5% and yours might be on 99.7% to accept someone as a trusted member of the community to interact with, but that makes our heads have to work harder and think more.
Which takes us to the problem or algorithms. How do we treat that negative information in terms of algorithms. eBay gives you an “average”, and that´s what makes you look at the 0.3%, but that might not be the most effective approach. Other algorithms could simply give more weight more weight to those negative experiences so that people can easily distinguish high reputation from bad reputation.
4. Reputation of criminals who trust each other.
Many criminals trust each other and they creat a network of trust between them. That’s a very complex problem to solve and needs to be looked after carefully in order to measure the diversity of reputation. You can have people transact 10 times on eBay with one another to earn “stars” and then once they have stars, commit some fraud (takes us back to the value of the source of reputation and Bill Gates recommendation).
5. People left out: offline to online reputation.
These people, that I mentioned earlier, like my mother. How do we make sure that they are not left out of opportunities? That they are also able to create those online transactions and to be part of this accessible network that we are creating? How do we extrapolate their reputation from the offline world into the online world to give everyone access to the economic graph?
6. Letting algorithms determine reputation. Inhumane?
Should we let numbers tell you what your reputation is? Or should we have full transparency in how the algorithm works? Or should we do it by hand? Let’s look at eBay again. The algorithm is only giving you the average but what you want is to actually see the list of transactions to find the bad ones, and that´s fine, there might be 3 bad ones out of 200. But that’s only eBay. What if there were 1 million transactions, for everything that you did in your life? Will you review all the raw data by hand? How do we even understand that in our brain? In my opinion we will probably need a mix of having access to the full history of transactions so that we can take a look at them individually, and then different companies will have a go to building the best algorithms who help people make sense out of that data from a design and metric point of view. But giving full power to algorithms to determine who is good to work with simply sounds wrong, and here I am talking more about the ethical and moral challenge than the technical challenge.
These are some of the challenges we are working on at Traity, which makes the subject fascinating. Would love to get your opinion, either here or happy to hear from you at firstname.lastname@example.org
Juan Cartagena, CEO
Originally published at blog.traity.com.